DOC23_ADDENDA_B_RT_ADJUDICATION_CARD_STAGED_RENDERED.md
Active Working and Red Team/DOC23 Working/DOC23 Red Teaming/Archived DOC23 Red Teaming/DOC23_ADDENDA_B_RT_ADJUDICATION_CARD_STAGED_RENDERED.md
ELNOR REPO READER TEXT MIRROR
Original path: Active Working and Red Team/DOC23 Working/DOC23 Red Teaming/Archived DOC23 Red Teaming/DOC23_ADDENDA_B_RT_ADJUDICATION_CARD_STAGED_RENDERED.md
Source repo: /Users/OpenClaw1/Elnor/Elnor Specs
Git branch: main
Git commit: dbaa25962edc11ab30e8d4ca1715f9ae5bf77331
Generated: 2026-06-09T01:23:58.539Z
---
# DOC23 Addenda B -- Red Team Adjudication Card (Staged Application Build)
**Status:** Draft, pending architect review + multi-LLM red-team. **Layers 0-5 complete (+ new rows)** -- front matter + Step 0 + Layer 1 (seven shared primitives) + Layer 2 (four flips, anchor-confirmed) + Layer 3 core (criticals/collisions/re-anchors/re-grades) + Layer 3 C-cluster math + standard rows. **Remaining:** Layer 4 (Sec. G as read-models + DOC20 OP-A), Layer 5 (fixtures), the new rows, then final pass. This is the assembled staged card; it supersedes `DOC23_ADDENDA_B_RT_ADJUDICATION_CARD_CONSOLIDATED.md` for application order (the consolidated card stays as the row-detail source).
**What this is.** The application-ready restructure of the 126-item, four-reviewer consolidated card, reorganized per the three reviews **of that card** (Claude + ChatGPT + Gemini). It supersedes the consolidated card for application order and for every item whose disposition or fix changed; unchanged CONFIRMED rows are carried by reference.
**Why staged (the central finding of all three card-reviews).** Do **not** apply the card as 126 flat rows. Apply a structural **Layer 1 first, defined once**, then the disposition flips, then the critical row fixes, then Sec. G, then fixtures. Defining shared primitives once removes the per-row redefinition that inflated the consolidated card and caused cross-row collisions.
```
Step 0 Strip / type-owner pass on the fix package itself (run BEFORE applying anything)
Layer 1 Shared structural primitives, defined once
Layer 2 Disposition flips (reverse/redirect prior rows) -- anchor-confirmed
Layer 3 Critical row fixes + cross-row collisions + re-grades + C-cluster math + standard rows
Layer 4 Sec. G Professional-Reliance / UX layer, as derived read-models (no new truth store)
Layer 5 Fixtures (golden scenarios + chaos/concurrency)
```
**Owner-first rule (binding; root cause of three reversed rows).** Before accepting any finding of the form "*X is missing / referenced but never defined*," resolve **whether X is already owned or defined elsewhere** (its owner doc, an existing enum, an existing field). A grep-level absence in one doc is not proof of absence in the system. Three consolidated-card rows (A-11, B-25, B-15) asserted a missing element that already exists or is owned elsewhere; they flip in Layer 2. Enforced mechanically in Step 0.
**Severity reframe (propagated everywhere).** A finding's severity is **importance** -- it drives revision priority and human-review surfacing -- and **never halts** the run. Only mechanical/system errors halt execution (`StepExecutionFailureKind`). An output is not auto-"passed" while serious findings are unresolved; if unresolvable, termination logic (V3.3.1 Sec. 6.7.3, see B-24) **escalates** (Gate-4 review or "rely with limitations"), it does not halt. Irreversible external actions (filing/sending) remain gated by an action permission -- an action gate, not a finding block. This retires the "blocking-authority guard" debate (A-05) and adjusts A-06 and B-24.
**Disposition vocabulary:** `ACCEPT` - `ACCEPT WITH MODIFICATIONS` - `ACCEPT-AS-FIX` - `DECLINE` (reverse a prior ACCEPT) - `REDIRECT-OP-A` (owned by another doc -> OP-A row) - `DEFER-PhaseB` (Sec. E) - `DECLINED` (Sec. F) - `ELEVATE` - `OPEN_FOR_ARCHITECT_REVIEW`.
**Coverage crosswalk.** Every one of the 126 consolidated-card items is accounted for in exactly one layer (the Layer Map). No item is dropped; flips and re-grades are explicit; Sec. E/Sec. F carry forward unchanged.
---
## Layer Map -- where each consolidated-card item lands
**Layer 2 -- Disposition flips / redirects** (all four anchor-confirmed -- see Layer 2)
- `A-11` -> **DECLINE** (`LearningMode` already enumerated, V3.3.1 Sec. 6.16.1); the *use* stays DEFER-PhaseB.
- `B-25` -> **DECLINE** (`depends_on_step_ids` + cycle detection Sec. 6.11.1 both exist); de-CRITICAL.
- `B-15` -> **REDIRECT-OP-A** (`RoomKind.plan_review` owned by DOC12, `OBL-DOC12-FORUM-01`).
- `D-04` -> **SPLIT**: keep Tier 0 (`lookup_receipt`), bar it from trust claims; retain the other three sub-fixes.
- `A-05` -> **subsumed by the severity reframe**; `A-06` verdict-mapping adjusts (standard row).
**Layer 3 -- Critical fixes, cross-row collisions, re-grades** (rendered in full):
`A-01` (->Layer 1) - `A-02` HIGH (elevation declined) - `A-04` (route-resolution) - `A-07` (one `NoVerdictReason` + aliases) - `A-12` (`OutcomeVerdict`) - `A-16` CRITICAL (`revision_in` chokepoint) - `A-23` (re-anchor to Source Workspace) - `A-26` (`FindingMatchKey`) - `B-02<->B-27` (`attempt_seq`+`prior_output_hash`) - `B-08` (dual `FailureKind` + `StepExecutionFailureKind`) - `B-24` CRITICAL (sufficiency predicate -> Sec. 6.7.3 loop breaker) - `B-26` (Sec. 6.11<->Sec. 6.7.3) - `C-03` (remove predicted hashes) - `D-01` (transitive taint) - `D-13` HIGH/injection.
**Layer 3 -- Math/scoring hardening** (C-cluster): `C-01`,`C-02`,`C-04`,`C-05`,`C-06`,`C-07`,`C-08`,`C-09`,`C-10` on the L1.5 shapes (`C-03` is in Layer 3 core). **`B-16` is a concurrency row, not a C-cluster item.**
**Layer 4 -- Sec. G Professional-Reliance / UX** (`G-01`...`G-21`): KEEP entire layer; build as **derived read-models** (`DerivedReadModelRecord`, Layer 1). UI surfaces (`G-07`, `G-12`...`G-16`, `G-20`) -> OP-A row to DOC20.
**Layer 5 -- Fixtures**: `G-21` (chaos/concurrency) + golden scenarios.
**Layer 3 -- Standard fixes (carried by reference):** all remaining Sec. A/Sec. B/Sec. C/Sec. D ACCEPT rows not named above -- ID + final disposition + consolidated-card anchor (full schema in the consolidated card).
**Sec. E (held, Phase-B)** `E-01...E-13` and **Sec. F (declined)** `F` -- carry forward unchanged; no per-row action.
**New rows added this restructure:** the Layer-1 primitives as first-class rows; reliance-decay; delegation; `D-19`/`D-20` completions; Grok Sec. 3.3 / Sec. 5.7 / Sec. 5.9. Citation fix: **B-24 is Grok Sec. 4.9**.
---
## Step 0 -- Strip / type-owner pass on the fix package (run before applying anything)
The **subtractive stroke applied to the fix package itself** -- removes adds that are redundant (already owned) or unspecifiable, and surfaces the Layer-2 flips. Runs once over the whole set before any row is applied.
1. **Admission gate (one-time, pre-write).** Every proposed type/enum/field/schema is checked against the `TypeOwnerRegistry` (Sec. L1.2). The gate is a **check, not a standing object**: passes only if the symbol is (a) new + assigned an owner doc, or (b) an extension of an owned symbol. New canonical object families trip `ARCHITECT_STOP`.
2. **Phantom-reference lint** (`SchemaReferenceValidationRule`, Sec. L1.2). Any reference to a field/type not in the registry fails. Catches "phantom schema refs" (A-09) at the package level. *This pass also caught two phantom fields in an early draft of this card -- `confidence_merge_strategy` and `tiebreaker_epsilon` -- both removed.*
3. **Owner-first resolution.** For every "X missing / referenced-never-defined" finding, confirm X's owner first. Output: the Layer-2 flips (A-11, B-25, B-15) + the D-04 hold.
4. **Collision scan.** Detect name collisions (e.g., `EvaluationVerdict` reused for an outcome concept; `FailureKind` defined twice). Output: Layer-3 renames (A-12->`OutcomeVerdict`, B-08->`StepExecutionFailureKind` + dual-`FailureKind` consolidation).
5. **Dedup-of-truth scan.** Detect rows inventing the same primitive under different names (termination counter, finding match key, read-model). Output: the Layer-1 consolidations (one `RevisorTerminationLedger`, one `FindingMatchKey`, the `DerivedReadModelRecord` family).
Step-0 lints:
```
step0.symbol_without_owner
step0.phantom_field_reference
step0.missing_claim_unresolved_owner // owner-first rule
step0.name_collision_across_package
step0.duplicate_primitive_across_rows
```
---
## Layer 1 -- Shared structural primitives (defined once)
Seven primitives defined **once** and referenced by ID from every later row. All three card-reviews built variants independently; consolidating them removes the per-row redefinition and the cross-row collisions. Each carries the rows it subsumes.
### L1.1 `GovernanceEnvelope` -- universal governance mixin
```ts
interface GovernanceEnvelope {
policy_decision_ref: PolicyDecisionRef // the EC decision that authorized this object's disposition
policy_generation_id: string // ties to the compiled policy version (B-12 freshness)
privilege_class: PrivilegeClass // attorney_client | work_product | confidential | none
access_tier: AccessTier // matter_team_access | supervising_attorney | firm_admin
taint_class: TaintClass // rides the object; consumers treat content as data, not instruction
matter_ref: MatterRef | null // scopes cross-matter visibility (D-24)
schema_version: 1
}
```
*Subsumes:* A-20, D-02, D-07/D-08, D-24. The Review Studio `LifecycleActorEnvelope` (DOC23_ADDB_Review_Studio.md Sec. 9.1) **extends** this mixin.
### L1.2 `TypeOwnerRegistry` + admission gate + phantom-reference lint
```ts
interface TypeOwnerRegistryEntry {
symbol: string; owner_doc: string; defined_at: string
status: "operative" | "proposed_this_pass" | "extension_of"
extends_symbol: string | null
schema_version: 1
}
// One-time admission check (Step 0, NOT a runtime object): pass iff symbol is (a) new + owner_doc, OR (b) extension_of an owned symbol. New canonical family -> ARCHITECT_STOP.
interface SchemaReferenceValidationRule { rule_id: "schema_reference_validation"; severity: "blocker" }
// FAIL: any field/type reference not present in TypeOwnerRegistry.
```
*Subsumes:* A-09, A-13 (dual learning envelope forbidden), A-25 (one owned name), D-22.
### L1.3 `EvaluationFinding` -- event / record / view (A-01, canonical)
Single finding model, shared with the Review Studio unit (DOC23_ADDB_Review_Studio.md Sec. 2). Replaces the two incompatible schemas; keeps **both** bases; the view is a read-only copy.
```ts
interface EvaluationFindingEvent { // immutable per-round audit record
finding_event_id: string; outcome_id: string; target_artifact_ref: string
target_version_ref: string // version-pinned (V3.3.1 Sec. 5.7)
target_scope_ref: TargetScopeRef | null // anchor; maps to a DOC20 CommentAnchor
detected_at: string; severity: FindingSeverity // IMPORTANCE, never a halt
assurance_basis: AssuranceBasis[] // user_instruction | saved_criteria | human_label | multi_reviewer_consensus | model_judgment_only
authority_basis: EvaluationAuthorityBasis[] // 9-value FD enum (FD Sec. 3.4) -- kept distinct from assurance_basis
match_key: FindingMatchKey // L1.4
body: string; governance: GovernanceEnvelope; schema_version: 1
}
interface EvaluationFindingRecord { // mutable lifecycle row
finding_id: string; origin_event_id: string
state: FindingState // active | resolved | human_resolved | superseded_by_revision | superseded_by_source_change | dismissed | contested
target_version_ref: string
resolved_by_receipt_ref: string | null // only a Revisor RevisionExecutionReceipt sets `resolved` (Sec. 5.7.2)
superseded_by_finding_id: string | null; governance: GovernanceEnvelope; schema_version: 1
}
interface FeedbackFindingView { // read-only projection the UI renders
finding_id: string; display_state: FindingState; anchor: TargetScopeRef | null
severity: FindingSeverity; body: string
history: FindingStateTransition[] // per-round snapshots -> cross-round drift view
schema_version: 1
}
```
*Subsumes:* A-01, A-05 (basis reconciliation; blocking part retired by the reframe), A-19 (`FindingState` negative exits), A-26 (`match_key`). `human_resolved` per DOC23_ADDB_Review_Studio.md Sec. 8.3.
### L1.4 `FindingMatchKey` + `RevisorTerminationLedger`
```ts
interface FindingMatchKey { // stable identity across revision rounds
artifact_root_id: string // stable root, NOT a version ref (A-26)
outcome_id: string; criterion_id: string | null
failure_signature: FailureKind // reuse the Sec. 6.2 FailureKind taxonomy (no new enum)
schema_version: 1
}
interface RevisorTerminationLedger { // ONE per revision run; all termination accounting here
run_id: string; rounds: RevisionRoundRecord[]
per_finding_attempts: { match_key: FindingMatchKey; attempt_seq: number; prior_output_hash: string }[] // B-02 <-> B-27
repeated_insufficiency_count: number; loop_breaker_limit: 3 // N = 3 (B-24)
sufficiency_predicate_results: SufficiencyPredicateResult[] // B-24 predicate feeding the Sec. 6.7.3 loop breaker
no_progress_detected: boolean
termination_reason: "converged" | "loop_breaker" | "escalated_gate4" | "rely_with_limitations" | null
schema_version: 1
}
```
*Subsumes:* A-26, B-24, B-02 <-> B-27, B-18/B-19, B-26 (one source of truth), D-18.
### L1.5 `FormulaEvaluationReceipt` + `MetricObservation` + `MetricRollup`
Makes every score a receipted formula evaluation, and gives the math-audit guards one home. Per-formula patches in Layer 3.
```ts
interface FormulaEvaluationReceipt { // one per score computation (C-01: Formula Registry)
formula_id: string // registered formula (versioned)
inputs: Record<string, number>
denominator_basis: string // e.g. "total_advice_rendered_count" not "accepted_count" (C-04)
sample_size: number
guards_applied: ("zero_denominator_fallback" | "underflow_clamp" | "percentile_bounded" | "censored_to_terminal")[]
output: number; schema_version: 1
}
interface MetricObservation {
metric_id: string; value: number
terminal_state_only: boolean // C-02: count only terminal-state cycles
observed_at: string; context_ref: string; schema_version: 1
}
interface MetricRollup {
metric_id: string; method: "mean" | "p50" | "p90" | "weighted_mean"
zero_denominator_fallback: "null" | "0" | "undefined_reported" // C-06
underflow_clamp: boolean // C-09: Math.max(0.0, weighted_mean - penalty)
percentile_impl: "rolling_window" | "t_digest" | "exact_bounded" // C-05: no unbounded array
missing_dimension_penalty_value: number // C-01
schema_version: 1
}
```
*Subsumes (shape for):* C-01 (Formula Registry -- keystone), C-02 (terminal-state censoring), C-04 (severity-weighted rate), C-05 (bounded percentile), C-06 (zero-denominator/insufficient-sample). C-07 (`CostEstimate`), C-08 (`NoveltyMetricSpec`), C-09 (task-mode/template-match) carry their own Layer-3 schemas but resolve `formula_id`s here. **(Corrections: `confidence_merge_strategy` and `tiebreaker_epsilon` removed -- both mis-anchored. A-04 is route-resolution (rendered in Layer 3); B-16 is a concurrency tiebreaker Sec. 11.9 (standard row), not a metric epsilon.)**
### L1.6 `DerivedReadModelRecord` + `ReadModelInvalidationSpec`
"No new truth store" primitive for Sec. G and cross-cutting surfaces (matches V3.3.1 Sec. 11.1.1 "derived, not invented").
```ts
interface DerivedReadModelRecord {
read_model_id: string
derived_from: string[] // source-of-truth refs; NEVER authoritative
projection_kind: string // task_health | findings_inbox | budget_narrative | run_diff | ...
invalidation: ReadModelInvalidationSpec
governance: GovernanceEnvelope // privilege/tier-filtered on read
schema_version: 1
}
interface ReadModelInvalidationSpec {
declared_dependencies: string[] // changing any invalidates the read-model (auditable, not predicted)
refresh: "on_dependency_change" | "on_read" | "scheduled"; schema_version: 1
}
```
*Subsumes:* Sec. G (G-01...G-21, Layer 4), the Review Studio read contract (Sec. 9.2), D-12.
### L1.7 `LintRegistry`
```ts
interface LintRegistryEntry {
lint_id: string; owner_doc: string; checks: string
severity: "advisory" | "warning" | "blocker"
waiver: { allowed: boolean; approver: "architect"; expires_at: string | null }
schema_version: 1
}
```
*Subsumes:* every scattered lint across Sec. A-Sec. D and Sec. G, plus the Step-0 and Review Studio lints.
---
## Layer 1 -- acceptance
Accepted when: (1) all seven primitives have an owner in the `TypeOwnerRegistry`; (2) no later row redefines them (referenced by ID only); (3) the Step-0 admission gate + phantom lint pass over the whole package; (4) the finding model matches the Review Studio unit; (5) `RevisorTerminationLedger` is the sole termination accumulator and the `FormulaEvaluationReceipt`/`MetricRollup` pair the sole score-computation shape.
---
## Layer 2 -- Disposition flips (anchor-confirmed)
Each flip was confirmed against **live spec text (section + line)** before being written, per the owner-first rule. These reverse or narrow prior ACCEPTs; the consolidated-card claim is quoted, then rebutted.
### L2.1 `A-11` -> **DECLINE** (enumeration redundant; the *use* remains DEFER-PhaseB)
- **Card claim (ACCEPT):** "`LearningMode` is referenced across Sec. 15 but never enumerated"; proposed `production | signal_generation | cross_calibration`.
- **Verified:** enumerated at **V3.3.1 Sec. 6.16.1** (3098-3108) as `"production" | "signal_generation" | "calibration"`; field `learning_mode: LearningMode` (3072); Sec. 0.4 inventory (45). The card looked in Sec. 15 and missed Sec. 6.16.1.
- **Disposition:** **DECLINE** the add -- it would create a divergent duplicate (`cross_calibration` != the real `calibration`). The behaviors (Sec. 6.16.2) remain **DEFER-PhaseB** (E-07).
- **Owner:** V3.3.1 Sec. 6.16.1.
### L2.2 `B-25` -> **DECLINE** (field *and* cycle-lint both exist; proposed remedy regresses the schema) -- was CRITICAL
- **Card claim (CRITICAL):** "Sec. 11.3's plan-lint references 'DAG acyclic' but `RevisionPlan` steps carry no dependency field." Proposed replacing `RevisionPlanStepBase`.
- **Verified:** `RevisionPlanStepBase.depends_on_step_ids: string[]` exists (3493); the check exists -- `RULE dag_acyclic` (4774-4775), Sec. 6.11.1 topological-sort-with-cycle-detection (2949), Sec. 11.22.1 (5833), `validation.dag_cyclic` (8175), deterministic plan linting includes DAG (1312).
- **Disposition:** **DECLINE**; de-escalate from CRITICAL. Both halves false. The proposed rewrite is **rejected** -- it would delete ~6 real fields (`step_order`, `affected_artifact_refs`, `affected_outcome_ids`, `source_repair_depth`, `idempotency_key`, `preconditions`). (B-26's `revalidation_policy` is handled on its own in Layer 3.)
- **Owner:** V3.3.1 Sec. 6.11.1 / Sec. 11.22.1.
### L2.3 `B-15` -> **REDIRECT-OP-A** (already owned by DOC12; the obligation already exists)
- **Card claim (ACCEPT, (!)verify):** "`RoomKind.plan_review` used but not registered."
- **Verified:** the Forum sub-addendum **itself** says DOC12 owns canonical room registrations incl. `RoomKind.plan_review` (45), the plan-review forum uses the DOC12-registered kind (754), config carries `room_kind?` resolving to DOC12 (179), and **`OBL-DOC12-FORUM-01`** already requires DOC12 to register it (874, 906-907).
- **Disposition:** **REDIRECT-OP-A.** Addenda B must not register it (would duplicate DOC12 ownership / violate the Step-0 admission gate). Already captured as `OBL-DOC12-FORUM-01`; carry on the OP-A list.
- **Owner:** DOC12 (`OBL-DOC12-FORUM-01`).
### L2.4 `D-04` -> **SPLIT**: keep Tier 0 (de-escalation hold) + retain the other three sub-fixes
- **Tier-0 sub-claim ("remove `0`") -> FLIP.** **Verified:** tier 0 = `lookup_receipt` is an intentional, load-bearing "receipt records only" tier for ephemeral lookups (SrcWS 103-106; "Check today's stock price -> tier 0", 239).
- **Disposition:** **KEEP Tier 0.** Reconcile the type to **include** 0 (`SourceRecord.tier in {0,1,2,3}`, 0 = `lookup_receipt`) rather than delete it. Add the safeguard: a Tier-0 `lookup_receipt` **must not** back a downstream reliance/trust claim -> `validation.tier0_receipt_used_as_trust_basis` (error).
- **Other three sub-fixes -> KEEP as ACCEPT** (standard rows): persist every `SourceTierTransition` (Sec. 3.3); align verification-state vocab (Sec. 2.3 / Sec. 4.1); gate demotion (`from_tier > to_tier`) by `cleared_by_access_tier >= "matter_team_access"` + reason (`validation.source_tier_demotion_without_authority`).
- **Owner:** Source Workspace Sec. 3 / Sec. 3.3.
### L2.5 `A-05` -> subsumed by the severity reframe (no separate fix)
The "blocking-authority guard" debate is retired by the reframe (findings carry importance, never halt; unresolvable serious findings escalate; irreversible external actions gated by an action permission). A-05's `assurance_basis` vs `authority_basis` content is captured in L1.3 (both bases retained).
### L2.6 New finding (surfaced by this verification): loop-breaker section citation is inconsistent
V3.3.1 cites the `D16` loop breaker as **both** Sec. 6.11 (1318, 1982, 9560) **and** Sec. 6.7.3 (2894, 3069, 3232) -- a concrete instance of **B-26**. Folded into B-26 (Layer 3).
### Layer 2 -- disposition delta & lints
Three items leave the ACCEPT-and-build set (`A-11` enum, `B-25`, `B-15`); `D-04` narrows; one CRITICAL retired (`B-25`).
```
layer2.declined_add_reintroduced
layer2.tier0_removed_from_domain
validation.tier0_receipt_used_as_trust_basis
```
---
## Layer 3 -- Critical fixes, collisions, and re-grades
These rows **change content**. Each verified against live spec before writing.
### L3.1 Critical / safety / compliance
**`A-16` -- Direct repair can bypass the central `revision_in` safety contract -- CRITICAL (confirmed).**
- **Verified:** `DirectFixStep` (V3.3.1 3560) is itself locked -- `target_port:"none_direct_fix"`, `direct_fix_class: DirectFixAllowedClass` (whitelist 601-608), `target_module_id`/`typed_instruction` FORBIDDEN, allow/forbid classes (601-617). The risk is the **FD wiring**: Feedback Delivery routes repair to `instruction_in`/`data_in`/`context_in` (FD 34), and the Revisor's safety chain runs through `revision_in` (FD 431). Those direct paths skip capability validation, preconditions, candidate versions, policy gates, preservation, and receipts.
- **Fix (FD Sec. 7.2):** repair may **execute** only through (a) a `revision_in` port, or (b) `revision_compatible = true` **and** `ModuleRevisionCapability` coverage; all other wiring is **advisory context only**.
```ts
interface PortRevisionEligibility { port_id: string; module_id: string;
revision_compatible: boolean; // default false
module_revision_capability_ref?: StorageRef; } // REQUIRED when revision_compatible = true
// validation.repair_routed_to_non_revision_port (error)
```
- **Cross-ref:** same chokepoint the Review Studio unit relies on (Sec. 1.2, Sec. 5.4). Pairs with A-17.
**`D-01` -- Transitive taint laundering -- compliance blocker (confirmed).**
- **Verified:** `SourceWorkspace.current_taint_class` exists (SrcWS 145), retrieval lineage exists (296-299), but there is **no `source_kind -> taint_class` map** and **taint is not transitive** -- a tainted artifact summarized by a sub-agent/forum re-enters "clean." Privilege/firewall-grade for a litigator.
- **Fix (SrcWS Sec. 4.1A + cross-addenda rule + `TaintAggregationPolicy`):** ship the default map (`web_source -> external_untrusted`; `case_law|statute|regulation|api_result|database_record -> external_authority_trusted`; `document|email|file|prior_task_output -> user_trusted_bounded`; `library_entry -> internal_corpus_trusted`); **transitive rule** -- any module/sub-agent output inherits the **max** taint of its inputs unless it passes a declared `SanitizationNode`; privileged input -> privileged workspace.
```ts
interface SanitizationNode { node_id: string; input_taint: TaintClass; output_taint: TaintClass;
method: "human_review" | "structural_strip" | "policy_filter"; evidence_ref: StorageRef; }
// validation.taint_not_inherited_through_summary (error); validation.source_kind_taint_unmapped (error)
```
- The `SanitizationNode` is the only sanctioned taint-downgrade. All taint rides `GovernanceEnvelope` (L1.1).
**`B-24` -- Revisor sufficiency protocol detection -- CRITICAL (confirmed; mechanism exists, detection underspecified).**
- **Verified:** the loop breaker exists -- `D16`, default **N=3** consecutive `still_failing_same_reason`, then BLOCK + escalate to Task Agent (Sec. 6.9) or human (Sec. 6.6) (1982, 2894, test F-LOOP-01 at 9080); `repeated_insufficiency_count` (3232). `still_failing_same_reason` = "same FailureKind + same primary finding" (1942). Underspecified: the **"same reason" detection**.
- **Fix:** the predicate is `FindingMatchKey` equality (L1.4); all accounting lives on the single `RevisorTerminationLedger` (L1.4). Hard Call surfaces only on a real user choice (missing source/verification/capability). The Sec. 6.11-vs-Sec. 6.7.3 split is fixed under B-26.
**`D-13` -- Cross-run `RunGuidance` persistence + contested-check -- ELEVATE to HIGH / injection-class (confirmed).**
- **Verified:** `RunGuidanceCandidate` (FD 296) / `RunGuidanceItem` (FD 311) exist with `guidance_kind in source_warning|do_not_use|must_include|style_rule` + `lifecycle_state`, but FD never says where guidance persists or how Run-A reaches Run B, `contested` is never checked, no lifecycle receipt.
- **Why elevated:** `must_include`/`do_not_use` is **injected** into later runs -- a cross-run injection vector; with D-01, tainted-derived guidance crossing runs is an injection path.
- **Fix:** persist `RunGuidanceItem` to `TaskBlueprint.persistent_guidance[]` (one authoritative store; EC durable write); `lifecycle_state: proposed|active|contested|superseded`, consumers MUST skip `contested`; emit `RunGuidanceLifecycleReceipt`; guidance carries `GovernanceEnvelope`. `validation.run_guidance_consumed_while_contested` (error). (Cross-run *learning* vs DOC72 precedence is Sec. E / E-02.)
### L3.2 Collisions / dedup
**`A-01`** -- finding model -> Layer 1 (L1.3). No separate fix.
**`A-26`** -- `ProgressSignal` keys on `(failure_kind, target_artifact_section_ref, finding_summary_hash)` (1948) -- version-sensitive. **Fix:** replace with `FindingMatchKey` (L1.4, stable `artifact_root_id`); `matched_findings` computed by key equality.
**`B-02 <-> B-27`** -- `regenerate` = "rerun from scratch with new instructions" (2739); B-02 under-specified idempotency, B-27 identical-output loop invisible. **Fix:** `RevisorTerminationLedger.per_finding_attempts` carries `attempt_seq` + `prior_output_hash` (L1.4); a regenerate matching `prior_output_hash` for the same `FindingMatchKey` counts as `still_failing_same_reason`.
**`B-08`** -- `FailureKind` is declared **twice** (Sec. 0.4.9 / 493 and Sec. 6.2 / 2715, same 12 values); `WorkspaceWriteFailureKind` (657) is separate. **Fix:** Sec. 6.2 canonical; Sec. 0.4.9 references it; step-execution failures -> `StepExecutionFailureKind` (distinct from evaluation `FailureKind` and `WorkspaceWriteFailureKind`). `supersession.dual_running_type_family`.
**`A-12`** -- Sec. 5.13 `EvaluationVerdict` is **per-lane** (`lane_*`, 2292 "Lane results aggregate into the OutcomeEvaluationResult"); no outcome-level type. **Fix:** add `OutcomeVerdict` (`passed|failed|indeterminate|not_applicable|blocked`); `EvaluationVerdict` stays per-lane; lane->outcome is the canonical mapping (CC Sec. 3.2).
**`A-07`** -- `evaluation_verdict` includes `indeterminate` (CC 204); `indeterminate_reasons: IndeterminateCause[]` cites a "V4 R203 taxonomy" (CC 214). **Fix:** one canonical `NoVerdictReason` + an **alias registry** mapping the variants; needs **no R203 knowledge**; full Addenda-A consolidation is an **OP-A row**.
### L3.3 Re-anchors / corrections
**`A-02`** -- Pattern C envelope field-location (ACCEPT/HIGH; **NOT elevated**). **Verified:** Sec. 3.3 requires `EvaluationResultEnvelope` be wrapped in `EvaluationArtifactEnvelope` (CC 289-295); the Judge needs `evaluated_target`/`evaluation_basis` but they live on `EvaluationFeedbackBundle`, so Pattern C breaks at read time. **Fix:** add both fields to `EvaluationResultEnvelope` (CC Sec. 3); bundle copy -> projection. **Re-grade:** high-importance correctness fix, no safety/compliance/data-loss dimension -> stays HIGH, not CRITICAL (correcting the early layer-map placeholder).
**`A-04`** -- Pattern C route-resolution policy (ACCEPT; corrected anchor). **Verified:** CC Sec. 3.7 resolution is "by consumer policy" with no schema. **Fix:** `EvaluationChainResolutionPolicy` + `DEFAULT_PATTERN_C_RESOLUTION_POLICY` (`qualitative_blockers_survive_numeric_pass: true`, `judge_can_override_evaluator: false`, `override_allowed_only_for_finding_states: ["contested","dismissed","rejected_by_user"]`, `route_precedence: "blocking_qualitative_first"`, `disagreement_route: "human_review"`). **Correction:** route-resolution, not confidence-merge -- the `confidence_merge_strategy` field mistakenly attached in L1.5 was removed.
**`A-23`** -- re-anchor `authority_level` to Source Workspace (corrected anchor). **Verified:** the card's `[FD Sec. 3/Sec. 5]` anchor is **wrong** -- `authority_level` lives in Source Workspace: `ApplicabilityScope.authority_level` (`binding|persuasive|advisory|unknown`, SrcWS 338) and `LegalSourcePayload.authority_level` (`binding|persuasive|distinguishable|adverse`, 359). **Fix:** reconcile the two: scope-level governs applicability; the domain payload (`distinguishable|adverse` = legal-specific extensions of `persuasive`) is descriptive. No FD change.
**`B-26`** -- standardize revalidation/termination naming. **Verified:** loop breaker cited as both Sec. 6.11 (1318, 1982, 9560) and Sec. 6.7.3 (2894, 3069, 3232). **Fix:** pick one anchor (recommend **Sec. 6.7.3**, where `repeated_insufficiency_count` lives, 3232); update the Sec. 6.11 cites. Do **not** introduce the invented `RevisionStepRevalidationPolicy` name; the policy is named on `RevisorTerminationLedger` (L1.4).
### L3.4 Re-grade summary
- `A-16` CRITICAL - `D-01` compliance blocker - `B-24` CRITICAL (mechanism exists, detection specified).
- `D-13` -> **HIGH / injection-class** (elevated).
- `A-02` -> **stays HIGH, not CRITICAL** (elevation declined on review).
- `B-25` -> **DECLINE, de-CRITICAL** (Layer 2).
- **Layer 1 corrections:** `confidence_merge_strategy` and `tiebreaker_epsilon` removed from L1.5 -- mis-anchored to A-04 (route-resolution) and B-16 (a concurrency row, Sec. 11.9) respectively.
### Layer 3 lints (registered in `LintRegistry`, L1.7)
```
validation.repair_routed_to_non_revision_port // A-16
validation.taint_not_inherited_through_summary // D-01
validation.source_kind_taint_unmapped // D-01
validation.run_guidance_consumed_while_contested // D-13
supersession.dual_running_type_family // B-08
progress_signal.match_on_version_sensitive_ref // A-26
```
---
## Layer 3 (continued) -- C-cluster math patches
All attach to the L1.5 primitives (`FormulaEvaluationReceipt` / `MetricObservation` / `MetricRollup`). `C-01` is the keystone -- the `FormulaRegistry` that L1.5 receipts reference; the rest are accepted reviewer patches (Gemini BUG-0x / ChatGPT CG-Mx), each verified against live spec before adoption.
**`C-01` -- Formula Registry (ACCEPT; keystone).** **Verified:** `QualityIndex` (V3.3.1 1600), `compiler_confidence_score` (1771, "0.0-1.0"), reputation/novelty/template-match/cost are all bare numbers with no formula; `unanchored_llm_judgment` is already excluded from `QualityIndex` aggregation (1600). **Fix:** the `FormulaRegistry` / `FormulaSpec` (typed inputs, output type/range/units, `missing_input_policy`, `zero_denominator_policy`, `normalization_policy`, `version`, `test_vectors`) + `CalibratedScore` + `QualityIndex` (aggregation_method, missing_dimension_policy) + `normalizeCriterionWeights()` (weights from criterion_weight | priority | uniform; `unanchored_llm_judgment` excluded). Owner: Common Contracts Sec. 9A (registry home, A-09). Every L1.5 `FormulaEvaluationReceipt.formula_id` resolves here. Lints: `validation.criterion_weight_sum_zero`, `validation.no_aggregation_eligible_criteria`. *Subsumes ~15 "this number has no formula" findings incl. D13 `compiler_confidence_score`.*
**`C-02` -- Survivorship bias in convergence metric (ACCEPT-AS-FIX).** **Verified:** `avg_revision_cycles_to_convergence` exists (V3.3.1 Sec. 15.1, line 7133); its denominator counts only outcomes reaching `satisfied`, erasing aborts/escalations. **Fix (Gemini BUG-01, verbatim):** bifurcate -- `avg_cycles_to_success` (denominator = outcomes transitioning to `satisfied`) and `wasted_cycle_burn_rate` (denominator = total cycles; numerator = cycles on outcomes that aborted/escalated/`unrecoverable`). Both are `MetricRollup`s with `terminal_state_only` observations (L1.5).
**`C-03` -- Hash-hallucination paradox (ACCEPT-AS-FIX).** **Already rendered in Layer 3 core** (predicted hashes verified at V3.3.1 3238-3239; remove `predicted_pre_hash`/`predicted_post_hash`; lock on stable section anchors; Dispatcher computes hashes at runtime). Listed here for cluster completeness.
**`C-04` -- Flat-ratio "weighted" reputation score (ACCEPT-AS-FIX; *use* -> Sec. E).** **Verified:** `advice_regression_rate` exists (V3.3.1 Sec. 15.8.2, line 7450); documented as 2x-weighted but computed flat; Phase-2 reputation-routing use is already deferred per Sec. 26.1 (7463). **Fix (Gemini BUG-03):** carry `severity_breakdown {minor,major,critical}`, `unweighted_regression_rate`, and `severity_weighted_penalty_score = ((minor*1)+(major*2)+(critical*5))/accepted_count`. The *consumer* (reputation gating routing) stays **DEFER-PhaseB** (E-06).
**`C-05` -- Latency mean skewed by thermal/cold-start outliers (ACCEPT-AS-FIX).** **Verified:** sub-agent/cost metrics live at Sec. 15.8 and Core Sec. 16.2; arithmetic latency mean is bimodal-unsafe on local Apple Silicon. **Fix (Gemini BUG-05):** `CostLatencyFinding` drops `avg_duration_ms`, carries `latency_distribution {p50_duration_ms, p90_duration_ms, sample_size}`. `MetricRollup.percentile_impl` (L1.5) supplies the bounded percentile computation. *(Note: the exact prior field name for the mean differs across Sec. 12.5/Sec. 16.2; the fix replaces any latency-mean with p50/p90 regardless of the field name.)*
**`C-06` -- Zero-denominator / insufficient-sample / cost tail-risk (ACCEPT).** **Fix:** every ratio metric declares `zero_denominator_policy: "undefined_insufficient_data"` (from C-01's `FormulaSpec`) and emits `CalibratedScore.sample_size`; below `min_sample_size` it returns `indeterminate`, not 0; `CostEstimate` (C-07) gains `p90_cost` beside `expected_cost`. Rides `MetricRollup.zero_denominator_fallback` (L1.5).
**`C-07` -- Shared `CostEstimate` / `TaskCostRecord` + metric-kind tag (ACCEPT WITH MODIFICATIONS).** **Fix:** canonical `CostEstimate {expected_cost, p90_cost, currency_or_token_unit, basis, sample_size}` + `TaskCostRecord` in Common Contracts Sec. 9A (never redefined per-doc); add `BudgetFailureKind` (`estimate_exceeded|hard_cap_hit|insufficient_budget_to_start`) and a `MetricKind` tag (`rate|count|calibrated_score|duration_distribution`) so every metric declares its kind.
**`C-08` -- Novelty metric space undefined (ACCEPT).** **Fix (Appx L):** `NoveltyMetricSpec` (feature-vector ref, distance metric in cosine/euclidean/jaccard/hybrid, range [0,1], `no_pattern_fallback`, calibration dataset, threshold) + `NoveltyAssessment` (closest pattern, distances, novelty_score, `forces_fresh_reasoning`, `triggers_task_agent`). A discriminated metric with an explicit no-pattern fallback -- no undefined space.
**`C-09` -- Task-mode / template-match calibration + magic 0.7 (ACCEPT).** **Fix (Appx K):** `TaskModeScoringFunction` (feature_weights, veto_precedence hard/soft, thresholds surfaced as config not a literal, `explicit_rule`) + `TemplateMatchFormula` (component_weights, `hard_veto_cap`, `soft_penalty_max_total`) + `computeTemplateOverallScore()` (weighted mean - soft penalties, capped on hard veto). Protects the direct-first default with a calibrated score, not a bare `0.7`. Lint: `validation.template_match_component_out_of_range`.
**`C-10` -- `criterion_semantics_hash` is lexical (ACCEPT).** **Fix:** rename to `criterion_text_hash` (lexical); add a separate `criterion_semantics_version` bumped only on a reviewed meaning change. Lint: `validation.criterion_semantics_version_unbumped_on_meaning_change` (advisory, human-reviewed).
### C-cluster lints (registered in `LintRegistry`, L1.7)
```
validation.criterion_weight_sum_zero // C-01
validation.no_aggregation_eligible_criteria // C-01
validation.template_match_component_out_of_range // C-09
validation.criterion_semantics_version_unbumped_on_meaning_change // C-10 (advisory)
```
---
## Layer 3 (continued) -- Standard ACCEPT rows (carried by reference)
These rows keep their consolidated-card disposition and fix verbatim -- the staged card carries them by **ID + disposition + consolidated-card anchor + one-line landing**, not re-rendered (the full schema lives in `DOC23_ADDENDA_B_RT_ADJUDICATION_CARD_CONSOLIDATED.md` at the cited section). Many land on a Layer-1 primitive (noted). None changed disposition.
**One row pulled up for emphasis -- `D-24` (privilege firewall, HIGH).** Cross-matter forum-post visibility (Run Board Sec. 5.4 / Source Workspace Sec. 9.4). **Lands on `GovernanceEnvelope.matter_ref` (L1.1):** a forum post / board event is visible only within its `matter_ref` unless an explicit cross-matter grant exists; `validation.cross_matter_forum_leak` (error). For a litigator this is privilege-grade -- high importance, though not a code-halt.
**Sec. A -- standard (17) (full bodies in Appendix R.1):**
| ID | Disp. | Anchor | Landing |
|---|---|---|---|
| A-03 | ACCEPT | CC Sec. 3.7 | Pattern C chain-ID lifecycle + chain registry (pairs A-25) |
| A-06 | ACCEPT | V3.3.1 Sec. 0.4.1 / CC Sec. 3.1-3.2 | `OutcomeEvaluationState` count & mapping; canonical mapping CC Sec. 3.2 (severity reframe note) |
| A-08 | ACCEPT | FD Sec. 6.2 | Full state->routing matrix; missing branches |
| A-09 | ACCEPT | CC Sec. 4.2 / V3.3.1 Sec. 5.17.7 | Phantom/misplaced refs -> handled by `TypeOwnerRegistry` (L1.2) |
| A-10 | ACCEPT-MOD | CC Sec. 3 / Sec. 11 | `EvaluationArtifactEnvelope` wrapper home; `overall_state` de-B-specced |
| A-13 | ACCEPT | Core Sec. 9 vs Common Sec. 5 | Core's duplicate learning envelope removed (registry forbids dual def, L1.2) |
| A-14 | ACCEPT | CC Sec. 4.2 | Qualitative-slice owner map; slice-required -> error not warning |
| A-15 | ACCEPT-MOD | CC Sec. 11.5 | "Build-ready" softened; command registry not mandatory |
| A-17 | ACCEPT | FD Sec. 5.3 / V3.3.1 Sec. 9 | Revisor is planner, not a revision target (pairs A-16) |
| A-18 | ACCEPT | FD Sec. 1.2 / Sec. 2.3 | Feedback-bundle emission discipline reconciled |
| A-19 | ACCEPT | V3.3.1 Sec. 0.4.4 / Sec. 5.7.1 | `FindingState` negative exits -> in L1.3 |
| A-20 | ACCEPT | CC Sec. 6.4 / Sec. 9.4 | `unanchored_llm_judgment` ack field -> `GovernanceEnvelope` (L1.1) |
| A-21 | ACCEPT | FD Sec. 2.2 / Sec. 3.3 | Optional-but-invariant-required fields (largely satisfied by A-01/L1.3) |
| A-22 | ACCEPT | CC Sec. 12 | Pending-consumer degraded-mode behavior |
| A-24 | ACCEPT | CC Sec. 3.4 vs V3.3.1 Sec. 5.18.8 | Parallel Judge+Evaluator example aligned to topology |
| A-25 | ACCEPT | CC Sec. 3.1 vs V3.3.1 Sec. 5.1 | `evaluation_chain_id` naming -> one owned name (fold A-03) |
| A-27 | ACCEPT | V3.3.1 Sec. 5.16 | `EvaluationSnapshot` source-workspace hashes typed correctly |
**Sec. B -- standard (20) (full bodies in Appendix R.2):**
| ID | Disp. | Anchor | Landing |
|---|---|---|---|
| B-01 | ACCEPT | V3.3.1 Sec. 11 | Unified `RevisionExecutionLifecycle` dispatcher<->receipts state machine |
| B-03 | ACCEPT (!)v | SrcWS / V3.3.1 Sec. 11.20 | Concurrent-write safety on Source Workspace |
| B-04 | ACCEPT | V3.3.1 Sec. 11.20.2 / Sec. 11.22 | Rolling hash under DAG/parallel (graph exists -- depends_on_step_ids, see B-25) |
| B-05 | ACCEPT | V3.3.1 Sec. 11.21 | Revalidation cascade convergence bound; de-dup upstream-failure rule |
| B-06 | ACCEPT | V3.3.1 Sec. dependency | Outcome-dependency cycle + `pending_dependency` deadlock |
| B-07 | ACCEPT | V3.3.1 Sec. 11.22 | Parallel sibling orphans after batch failure; parallelism first-class |
| B-09 | ACCEPT | V3.3.1 Sec. candidate | Candidate external side-effects; lifecycle vs head-pointer |
| B-10 | ACCEPT | V3.3.1 Sec. 11 / FD | `TaskCancelProtocol` + Hard Call blocking scope |
| B-11 | ACCEPT | V3.3.1 Sec. skip / FD | `TaskSkipProtocol` + skip receipts |
| B-12 | ACCEPT | CC Sec. policy | Policy-freshness field -> `GovernanceEnvelope.policy_generation_id` (L1.1) |
| B-13 | ACCEPT | SrcWS Sec. 6 | `ResearchNeed` lease |
| B-14 | ACCEPT | Task Forum | Forum deadlock breaker (+ remove `task_agent_decides` per ChatGPT) |
| B-16 | ACCEPT-AS-FIX | V3.3.1 Sec. 11.9 | **Concurrency** tiebreaker: replace wall-clock `created_at` with deterministic ordering (NOT a metric epsilon) |
| B-17 | ACCEPT | FD Sec. 8.4 / Sec. 10.3 | `instruction_in` overload -> DOC23/DOC15/DOC24 boundary |
| B-18 | ACCEPT | V3.3.1 Sec. 6.5.2 / Sec. 7.9 | `HardRevisionCall.options` bounded-nonempty -> on `RevisorTerminationLedger` (L1.4) |
| B-19 | ACCEPT | V3.3.1 Sec. 11.6 / Sec. 0.4.7 | `hard_call_resolved` producer -> L1.4 |
| B-20 | ACCEPT | V3.3.1 Sec. 6.7.2 / Sec. 11.21 | Success-condition 5 races cascade |
| B-21 | ACCEPT | Core Sec. 9.0.6 | Signal-emission ordering vs receipt persistence |
| B-22 | ACCEPT | V3.3.1 Sec. 5.18 / Sec. 11.15 | Pattern C latency/cost budget |
| B-23 | ACCEPT | FD Sec. 6.3 | Multiple delivery branches idempotency |
**Sec. D -- standard (20; D-24 pulled up above) (full bodies in Appendix R.4):**
| ID | Disp. | Anchor | Landing |
|---|---|---|---|
| D-02 | ACCEPT | SrcWS Sec. 2 | Workspace taint aggregation -> `GovernanceEnvelope` + max-taint (D-01) |
| D-03 | ACCEPT | SrcWS Sec. 0.3 / V3.3.1 Sec. 12 | `TaskSourceWorkspace` vs `SourceWorkspace` identity |
| D-05 | ACCEPT | SrcWS Sec. 6.2 / Sec. 7.4 | `ResearchNeed` scoping, ref types, `human_needed` exit |
| D-06 | ACCEPT | SrcWS Sec. 4 / Sec. 5 / Sec. 6.3 | Evidence anchors, payload registry, tool-receipts, extractor-vs-source |
| D-07 | ACCEPT | Task Forum Sec. 5.2 / Sec. 5.3 | Forum post visibility/supersession/governance -> `GovernanceEnvelope` (L1.1) |
| D-08 | ACCEPT | Task Forum Sec. 6.3 / Sec. 6.4 | Context-packet ownership, omission manifest, audience enum, budget |
| D-09 | ACCEPT | Task Forum Sec. 1 / Sec. 5 | Passive-board auto-publish privacy/volume controls |
| D-10 | ACCEPT | Task Forum Sec. 7.2 etc. | `ModuleAssistanceRequest`, participant/moderator model, moderator-failure |
| D-11 | ACCEPT | Run Board Sec. 3.1 / Sec. 6.2 | `BoardDigest` filter rule |
| D-12 | ACCEPT | Run Board Sec. retention | Retention/compaction/event-class -> `DerivedReadModelRecord` (L1.6) |
| D-14 | ACCEPT | Core Sec. 3D / Sec. 4B / Sec. 5A | `InjectionSlotRegistry`, compact cards, command registry, token_budget, DOC24 ownership |
| D-15 | ACCEPT | V3.3.1 Sec. 8.4 / Sec. 17.1 | Sub-agent output-contract plurality, no-sub-agent fallback |
| D-16 | ACCEPT (!)v | SrcWS / Forum | Workspace API operations defined |
| D-17 | ACCEPT | CC Sec. 7 / V3.3.1 Sec. 7.9.3 | Anchor/hash hygiene: empty `StructuredAnchor`, uncomputed `context_hash` |
| D-18 | ACCEPT | V3.3.1 Sec. repeated-failure | Repeated-failure keyed on versioned refs -> `FindingMatchKey` (L1.4) |
| D-19 | ACCEPT | SrcWS Sec. SourceRecord | Per-module cost attribution field -> `CostEstimate`/`TaskCostRecord` (C-07) |
| D-20 | ACCEPT | SrcWS Sec. library | Library-promotion gate EC policy defined |
| D-21 | ACCEPT | V3.3.1 / Core | `requires_background_progress` overload split |
| D-22 | ACCEPT-MOD | CC Sec. 11.5 | Backward-compat softened; section-anchor hygiene |
| D-23 | ACCEPT | FD Sec. 9.4 | "Silent ignoring fires validation" made enforceable |
**Sec. E (held, Phase-B) E-01...E-13** and **Sec. F (declined) x4** -- carry forward unchanged; no per-row action. **Sec. G (G-01...G-21)** -> Layer 4 (next).
---
## Layer 4 -- Sec. G Professional-Reliance / UX layer
Sec. G is the differentiator (pre-flight, reviewable diffs, evidence binder, plain-English cost, a single "can I rely on this" memo). All 21 rows KEEP their ACCEPT disposition. The staged-card value-add is to **classify** them correctly against the "no new truth store" rule -- because the reviews' "build Sec. G as read-models" is true for the *surfaces* but wrong for a handful of items that carry genuine new state or are operations. Three kinds:
### L4.1 Reliance artifacts (persisted; mostly derived, a few new-truth fields)
Schemas already written in the consolidated card (`[Appx O]`/`[Appx M]`); carried by reference here with classification + governance. All carry `GovernanceEnvelope` (L1.1) and are matter-scoped (D-24).
- **`G-06` `TaskReliancePacket` -- capstone (ACCEPT).** [CC/Core, Appx O] Assembled from existing truth (evaluation_chain_ids, evidence package, hard-call resolutions, policy decisions, budget narrative, known-good states) -> **derived**; the only computed new value is `reliance_status in safe_to_rely_within_scope | rely_with_limitations | not_safe_to_rely | human_review_required` + `reliance_scope`. `unresolved_limitations` uses the A-07 limitation taxonomy. Everything else in Sec. G feeds it.
- **`G-01` `EvaluationContractReview` (ACCEPT).** [Appx O] Pre-execution; interpreted goal/criteria/thresholds are **derived** from the compiled plan, but `approval_status` (`pending|approved|rejected|edited|waived_by_policy`) is **genuine new truth** (the user's decision) -> persisted artifact, not a pure read-model. Gate optional per task autonomy.
- **`G-02` `RevisionReviewPacket` (ACCEPT).** [Appx O] before->candidate semantic diff + `finding_to_change_map` + preservation/revalidation results are **derived**; `reviewer_action` (`accept|reject|fork|request_changes|restore_known_good_state|no_user_review_required`) is **new truth**. Pairs with A-16 (revision flows through `revision_in`) and G-04 (restore).
- **`G-03` `EvidencePackage` (ACCEPT).** [Appx O] **Derived** export (snapshot + `claim_support_map` over D-06 evidence anchors); persisted as the binder. No new truth beyond the snapshot boundary.
- **`G-05` `BudgetNarrative` (ACCEPT).** [Appx O] **Derived** from `TaskCostRecord`s (C-07); separates logical LLM calls from infrastructure retries. Feeds G-09.
- **`G-07` `AttentionLedger` / `DecisionQueue` (ACCEPT -- author minimal).** Cross-run queue of pending hard calls / blocked / contested / approvals. **Derived** projection over those items, but each `AttentionLedgerItem` carries a mutable `resolved_at` -> persisted queue. UI surface -> DOC20 (L4.4).
### L4.2 Operations / state primitives (NOT read-models)
These are state or actions, not projections -- they need real schemas/commands.
- **`G-04` `KnownGoodState` (ACCEPT).** [Appx M] Named restorable checkpoint (artifact_version_refs[] + workspace_snapshot_ref + label). A saved **pointer set**, plus a `restore` **command** in the registry (D-14) with an idempotency key. Backs B-10 cancel, G-17 fork, G-02 restore.
- **`G-17` `TaskRunFork` (ACCEPT).** Fork a run from a `KnownGoodState`; the honest form of the declined ShadowWorkspace (Sec. F-01) -- carries `irrevocable_side_effects_at_fork[]` (an already-sent email stays sent). An **operation** + record.
- **`G-18` `ExplanationTrace` (ACCEPT).** Short human-readable causal trace emitted by every `CompiledRevisionStrategy`/`RevisionPlan` ("changed X because finding Y, preserving Z") -> a small **new artifact**, `explanation_trace_ref` on the plan. Feeds G-15.
- **`G-19` `TaskReplay` (ACCEPT).** Deterministic replay primitive keyed to a run snapshot + `KnownGoodState`. An **operation**; underpins G-14/G-16.
### L4.3 DOC20 read-model surfaces (`DerivedReadModelRecord`, L1.6 -- no new truth)
Each is a projection over existing truth; none is authoritative. Rendered as `DerivedReadModelRecord` with the noted `projection_kind` / `derived_from`. All matter-scoped (D-24); contestable items carry the G-11 contest affordance.
| Row | `projection_kind` | `derived_from` |
|---|---|---|
| `G-12` `WorkProductCertification` (highest-leverage) | `work_product_certification` | `TaskReliancePacket` (G-06) |
| `G-13` `FindingsInbox` | `findings_inbox` | canonical `EvaluationFinding`s (A-01/L1.3); matter-scoped (D-24) |
| `G-14` `RunDiff` | `run_diff` | run records + `BudgetNarrative` (G-05) |
| `G-15` `DecisionAuditView` | `decision_audit` | Pattern C chain (A-03) + routing (A-08) + the coordination trace |
| `G-16` `RunReplayPreview` | `run_replay_preview` | `TaskReplay` (G-19) + `KnownGoodState` (G-04) |
| `G-20` Unified Evaluation-Chain view | `evaluation_chain` | the chain registry (A-03); qualitative + quantitative side by side |
| `G-08` `TaskHealthCard` | `task_health` | run state + `BudgetNarrative` (G-05) + last `OutcomeEvaluationState` |
| `G-09` cost forecast (pre-run) | `cost_forecast` | shared `CostEstimate` (C-07) summed over packets/research/LLM (D-18 sub-totals); surfaced in G-01, reconciled by G-05 |
| `G-10` unified review entry | `review_home` | composes G-13 (triage) + G-02 (changes) + G-15 (why) |
| `G-11` contestability surfacing | `contestable_marker` | defeasible findings (FD Sec. 3.4); marks contestable findings/verdicts inline |
`G-08`/`G-10`/`G-11` are pure surfaces; `G-12`-`G-16`/`G-20` are user-facing surfaces over the artifacts above.
### L4.4 OP-A row -- DOC20 UI surfaces
The contracts above are defined here (MOCKUP-READY: contracts tight, UI light); the **UI rendering is owned by DOC20**. One OP-A row:
```
OP-A -> DOC20: render the Sec. G surfaces -- WorkProductCertification (G-12), FindingsInbox (G-13), RunDiff (G-14),
DecisionAuditView (G-15), RunReplayPreview (G-16), Unified Evaluation-Chain view (G-20), TaskHealthCard (G-08),
AttentionLedger/DecisionQueue (G-07), unified review home (G-10), contestability markers (G-11).
All as DerivedReadModelRecord-backed surfaces (no new truth); matter-scoped per D-24; privilege/tier-filtered on read.
```
### G-21 -> Layer 5
Chaos/concurrency fixtures (storage-full, malformed LLM output, mid-run privilege change, clock skew, parallel writes) back B-03/B-04/B-05/B-16 and D-01/D-24 -- rendered in **Layer 5** (fixtures), where the runtime fixes gate on them passing.
### Layer 4 lints (registered in `LintRegistry`, L1.7)
```
reliance.packet_status_without_evidence_or_limitations // G-06: reliance_status set with neither evidence refs nor a limitation list
read_model.authoritative_write_attempt // L1.6: any Sec. G surface attempting a write to source-of-truth
ui.contestable_finding_without_affordance // G-11: a defeasible finding rendered without the contest control
attention_ledger.item_without_command_or_resolution // G-07: a pending item with no command_ref and no resolved_at path
```
---
## Layer 5 -- Fixtures (golden scenarios + chaos/concurrency)
Golden scenarios are **executable assertions**, not names. Each backs a specific row and states setup -> expected outcome -> the lint/gate it exercises. Fixture shape:
```ts
interface FixtureRecord {
fixture_id: string; backs_rows: string[]; setup: string; expected: string;
asserts_lint?: string; gate: "slice_local" | "cross_layer" | "final_switchover";
blocking_severity: "advisory" | "warning" | "blocker"; schema_version: 1;
}
```
### L5.1 Golden scenarios (correctness fixes)
| ID | Backs | Setup -> Expected | Lint |
|---|---|---|---|
| GS-01 | A-16 | A repair `OutcomeRepairInstruction` is wired to `instruction_in`/`data_in`/`context_in` -> it executes **no** mutation; carried as advisory context; only a `revision_in` (or `revision_compatible=true` + capability) port mutates | `validation.repair_routed_to_non_revision_port` |
| GS-02 | D-01 | A `web_source` (`external_untrusted`) artifact is summarized by a sub-agent and consumed by the Revisor -> the summary inherits `external_untrusted`; unless it passed a `SanitizationNode`, the Revisor treats it as data, not instruction | `validation.taint_not_inherited_through_summary`, `validation.source_kind_taint_unmapped` |
| GS-03 | B-24 | Revisor produces 3 consecutive outputs with the **same `FindingMatchKey`** unresolved -> the 4th attempt is BLOCKED; `RevisorTerminationLedger.termination_reason = "loop_breaker"`; escalates per Sec. 6.7.3 | (extends F-LOOP-01) |
| GS-04 | D-13 | A `RunGuidanceItem` in `lifecycle_state: contested` is present at run start -> every consumer skips it; a `RunGuidanceLifecycleReceipt` records the skip | `validation.run_guidance_consumed_while_contested` |
| GS-05 | D-04 | A Tier-0 `lookup_receipt` is cited as support for a reliance/trust claim -> error; Tier 0 remains valid as a "lookup occurred" record | `validation.tier0_receipt_used_as_trust_basis` |
| GS-06 | D-24 | A forum post with `matter_ref = M1` is requested by a reader scoped to `M2` with no cross-matter grant -> blocked | `validation.cross_matter_forum_leak` |
| GS-07 | A-02 | Pattern C: the Judge reads `evaluated_target`/`evaluation_basis` -> present on `EvaluationResultEnvelope` (post-fix); chain resolves at read time | (read-time resolution) |
| GS-08 | B-08 | Build-time scan of V3.3.1 -> exactly **one** `FailureKind` definition (Sec. 6.2 canonical; Sec. 0.4.9 references it); step-execution failures use `StepExecutionFailureKind` | `supersession.dual_running_type_family` |
| GS-09 | C-03 | A multi-step `rolling_hash_in_place` plan is compiled -> the LLM emits **no** `predicted_post_hash`; the Dispatcher locks on `target_section_anchor_hash` and computes hashes at runtime; the plan does not crash | (no predicted hash in output contract) |
| GS-10 | C-01/C-06 | A ratio metric runs with a zero denominator -> returns `indeterminate` (not 0/NaN), `CalibratedScore.sample_size` set; every score resolves a `FormulaSpec.formula_id` | `validation.criterion_weight_sum_zero` |
| GS-11 | G-06 | A `TaskReliancePacket` sets `reliance_status` -> it carries either evidence refs or a non-empty `unresolved_limitations`; a packet with neither fails | `reliance.packet_status_without_evidence_or_limitations` |
| GS-12 | A-12 | A lane returns `lane_failed` while the outcome is otherwise satisfied -> `OutcomeVerdict` is computed by the canonical lane->outcome mapping, not read off a single `EvaluationVerdict` | (verdict-level separation) |
### L5.2 Chaos / concurrency fixtures (`G-21` -- the Sec. B harness)
The runtime fixes **gate** on these passing (blocking).
| ID | Backs | Setup -> Expected |
|---|---|---|
| CH-01 | B-03/B-09 | Storage fills mid-write -> `WorkspaceWriteFailureKind` (`partial_artifact_written`); the head pointer is not advanced to a non-existent artifact; no orphan |
| CH-02 | B-08 | The Revision Compiler emits malformed/invalid plan JSON -> classified as a `StepExecutionFailureKind` and **halts mechanically** (not surfaced as a low-severity finding) |
| CH-03 | D-01/D-24 | Privilege downgrades mid-run (matter access revoked) -> `GovernanceEnvelope` re-evaluated against the new `policy_generation_id`; in-flight reads fail-closed |
| CH-04 | B-16 | Two events carry equal/inverted wall-clock `created_at` -> a deterministic tiebreaker (not `created_at`) fixes a stable order |
| CH-05 | B-04/B-07 | Two revision steps target one artifact under the DAG -> the precondition/rolling-hash check rejects the stale write; no lost update; parallel siblings not orphaned on one failure |
| CH-06 | B-05 | A revalidation cascade is triggered -> it converges within the declared bound; the duplicated upstream-failure rule fires once |
| CH-07 | B-06 | A `pending_dependency` cycle is introduced -> detected by `dag_acyclic` / topological cycle detection (B-25 verified mechanism); no deadlock |
### L5.3 Fixture-to-row matrix & gate
- **slice_local:** GS-07, GS-08, GS-10, GS-12. **cross_layer:** GS-01...GS-06, GS-09, GS-11, CH-01...CH-07.
- **Gate:** the runtime/concurrency fixes (B-03/B-04/B-05/B-06/B-07/B-09/B-16) and the compliance fixes (D-01/D-24) **do not ship** until CH-01...CH-07 pass; the safety fix (A-16) and taint fix (D-01) gate on GS-01/GS-02.
- All fixtures register their lints in `LintRegistry` (L1.7); none introduces new truth.
---
## New rows (added by this restructure)
Rows the restructure introduces or completes beyond the 126 consolidated-card items. Each traces to an existing decision or a named primitive; none is a free invention.
### NR.1 Layer-1 primitives as first-class rows (`TypeOwnerRegistry` entries)
The seven primitives get tracked rows so their ownership is explicit (not floating):
| Primitive | Owner doc | Status |
|---|---|---|
| `GovernanceEnvelope` (L1.1) | Common Contracts | new mixin |
| `TypeOwnerRegistry` + `SchemaReferenceValidationRule` (L1.2) | Common Contracts (governance) | new |
| `EvaluationFinding` event/record/view (L1.3) | Common Contracts Sec. 4.2 (A-01); shared w/ Review Studio Sec. 2 | replaces two schemas |
| `FindingMatchKey` (L1.4) | V3.3.1 (reuses Sec. 6.2 `FailureKind`) | new |
| `RevisorTerminationLedger` (L1.4) | V3.3.1 Sec. 6.7.3 | new (consolidates counters) |
| `FormulaEvaluationReceipt`/`MetricObservation`/`MetricRollup` (L1.5) | Common Contracts Sec. 9A | new (C-01 registry) |
| `DerivedReadModelRecord`/`ReadModelInvalidationSpec` (L1.6) | DOC20 / Common Contracts | new |
| `LintRegistry` (L1.7) | Common Contracts (governance) | new |
### NR.2 Reliance-decay (new)
Reliance is time- and dependency-bounded. **Fix:** `TaskReliancePacket` (G-06) gains `valid_until: ISO8601` + `decay_policy`; a `RelianceDecayCheck` (a `DerivedReadModelRecord`, L1.6, whose `declared_dependencies` are the packet's source/finding/policy refs) downgrades `reliance_status -> "human_review_required"` when any underlying source/finding/policy has changed or aged past threshold. Lint: `reliance.packet_consumed_past_valid_until`.
### NR.3 Delegation (new)
A Hard Call or scoped task can be delegated (e.g., to a supervising attorney, or a sub-agent within scope). **Fix:**
```ts
interface DelegationGrant {
grant_id: string; delegator_ref: string; delegate_ref: string;
scope: { decision_kinds: HardRevisionCallKind[]; outcome_ids?: string[]; matter_ref?: MatterRef };
authority_basis: EvaluationAuthorityBasis; expires_at: ISO8601; revocation_ref?: string;
governance: GovernanceEnvelope; schema_version: 1;
}
// A delegated decision carries the delegator's access_tier; a delegated Hard Call resolution records grant_id.
```
Ties to `access_tier` (e.g., `supervising_attorney`) and surfaces in the `AttentionLedger` (G-07). Lints: `delegation.decision_outside_granted_scope`, `delegation.expired_grant_used`.
### NR.4 `D-19` / `D-20` completions (complete two Sec. D standard rows)
- **`D-19`** (per-module cost attribution): add `cost_attribution_ref: TaskCostRecord` (C-07) on `SourceRecord` / module output, so cost rolls up per module. Closes the "no field" gap.
- **`D-20`** (library promotion gate): define `LibraryPromotionPolicy` -- `promote_to_doc73_library_candidate` requires an EC `PolicyDecisionRef` (privilege/matter scope) **and** source tier >= 2 before promotion to the DOC73 library. Lint: `library.promotion_without_ec_policy`.
### NR.5 Grok Sec. 3.3 -- feedback->prompt responsibility matrix
**Fix:** `FeedbackResponsibilityMatrix` maps each feedback kind -> the component responsible for turning it into the next instruction: repair instruction -> Revisor (via `revision_in`, A-16); strategic/scope change -> Task Agent; ambiguous/contested -> human (Gate-4). Prevents feedback from being dropped or double-handled. Lint: `feedback.without_responsible_owner`.
### NR.6 Grok Sec. 5.7 -- sub-agent metric observation
**Fix:** sub-agents MUST emit `MetricObservation`s (L1.5) for cost/latency/outcome so metrics aggregate across the run; without it sub-agent work is invisible to the C-07 cost rollup and the Sec. 15 quality metrics. Contract: every sub-agent activation emits >=1 `MetricObservation` keyed to its activation_seq. Lint: `subagent.work_without_metric_observation`.
### NR.7 Grok Sec. 5.9 -- BoardDigest ResearchNeed import
**Fix:** the Run Board's `BoardDigest` (D-11) imports open `ResearchNeed`s (D-05/D-13) as a `research_needs_open[]` projection (a `DerivedReadModelRecord`, matter-scoped per D-24) so research gaps surface on the board rather than staying buried in the workspace.
### NR -- citation/scope fixes carried
- **B-24 is Grok Sec. 4.9** (not Sec. 4.10).
- **B-14** also removes `task_agent_decides` (ChatGPT) -- folded into the B-14 standard row.
---
## R0.4 audit additions (completeness-audit fold-ins, new rows, OP-A, parked, confirmed divergences)
A source-based completeness audit of this card against the three card-reviews (1,182 extracted findings; per-row cross-check plus a deterministic name/lint check against **both** this card and the consolidated card) confirmed the structural layer was faithful and surfaced a detail layer the card-reviews added atop the consolidated card. Accepted results are folded in here and rendered inline in Appendix R.
### Audit fold-ins (R0.4 deepening) -- rendered inline in Appendix R
| # | Row | Fold-in | In |
|---|---|---|---|
| 1 | C-07 | `CostVector` / `CostDimension` / `CostCorrelationPolicy` (quantile-summing) | R.3 |
| 2 | G-04 | `KnownGoodCheckpoint` hardening (policy + capability snapshot; irreversible-restore lint) | R.7 |
| 3 | G-02 | `RevisionReviewDecisionReceipt` + review lints (resolves the inline-action divergence) | R.7 |
| 4 | G-19 | `TaskReplayRequest` / `ReplayDivergenceRecord` + replay lints | R.7 |
| 5 | A-02/A-04 | `EvaluationChainResolutionReceipt` | R.1 |
| 6 | B-16 | `PlanSelectionTiebreakPolicy` (epsilon band on `plan_risk_score`) + concurrency lints -- corrects the `tiebreaker_epsilon` removal | R.2 |
| 7 | B-03 | `SourceWorkspaceOperation`(+`Receipt`) write-precondition / lock-lease API | R.2 |
| 8 | B-05/B-06 | `RevalidationCascadeRun` + `OutcomeDependencyGraphPolicy` (bounded cascade) | R.2 |
| 9 | D-06 | `ClaimSupportDerivationPolicy`/`Receipt` (+ `SourceEvidenceAnchor`) | R.4 |
### New rows (in-scope, R0.4)
**NR.8 -- `ReceiptCoverageRegistry`** -- which operations MUST emit a receipt; closes the gap where receipts exist per-feature but nothing enforces coverage.
```ts
interface ReceiptCoverageRegistryEntry {
operation_kind: string; // "revision_step" | "workspace_op" | "review_decision" | "formula_eval" | ...
required_receipt_type: string; // the receipt schema that must be emitted
enforcement: "blocking" | "warning";
}
// lint: validation.operation_without_required_receipt
```
**NR.9 -- `PlanVerifierRequirementPolicy`** -- when a `RevisionPlan` must pass verification before execution.
```ts
interface PlanVerifierRequirementPolicy {
requires_verification_when: ("multi_step" | "irreversible_effect" | "in_place_mutation" | "cross_artifact")[];
verifier: "dag_acyclic_plus_precondition" | "human_gate_4";
on_verification_fail: "block" | "downgrade_to_advisory";
}
// lint: validation.plan_executed_without_required_verification
```
### Cross-doc obligation (OP-A)
**OBL-DOC24-CTXPKT-01** -- `ContextPacketFidelityContract` + `TaskKnowledgePackFreshnessPolicy`: the context packet a task run receives must declare a fidelity contract (what was included/omitted vs. the source set) and a freshness policy (when a knowledge pack is stale and must be re-pulled). **Owner: DOC24** (context assembly/delivery); referenced by Addenda B at the task-run boundary. Routed to OP-A -- not rendered as a DOC23 schema. DOC24-owned lints: `validation.context_packet_without_fidelity_contract`, `validation.task_knowledge_pack_consumed_past_freshness`.
### Parked -- OPEN_FOR_ARCHITECT_REVIEW (in/out ruling pending; not operative)
- **`WorkspaceExternalizationPolicy`/`Receipt`** -- externalize large workspace contents to storage under a cost cap.
- **`ExternalSourceQueryPolicy`/`Receipt`** -- govern external (web/API) source queries (authority signal, rate, receipt).
- **`TaskAgentForumSurfaceOwnership`** -- who owns the forum surface a task agent posts to (may fold into B-15 / DOC12).
### Confirmed divergences (reviewed; intended)
- **A-02 -> HIGH, not CRITICAL** (2-3 reviewers said CRITICAL): a correctness fix with no safety/compliance/data-loss dimension; the silent-wrong-verdict concern is covered by the escalation path plus the new A-02/A-04 chain-resolution receipt.
- **A-05 blocking reframed**: both basis arrays kept (closes the lossy-collapse build-blocker); the blocking role moves to severity + an action-permission gate on publishing/sending, not a typed `BlockingAuthorityEvaluation`.
- **Row-merges declined**: A-08+B-23 and B-05+B-06+B-20 stay distinct rows (shared mechanisms rendered; row identity preserved for traceability).
- **Admission gate** is a one-time pre-write `TypeOwnerRegistry` check, not a standing object.
---
## Appendix R -- Full row renderings (self-contained reference)
This appendix renders every adjudicated row in full so the card is self-contained for red-team and for drafting the next spec version in a fresh context. Rows restructured in the Layers above show a pointer (the Layer rendering governs); standard / held / declined / Sec.G rows are rendered in full from the adjudicated bodies. The nine audit fold-ins (R0.4) surfaced by the completeness audit are marked inline.
### R.1 -- Sec.A rows
**A-01** -> rendered in Layer 1 (that rendering governs).
**A-02** -> rendered in Layer 3 core (that rendering governs).
**Audit fold-in (R0.4) -- chain-resolution receipt** (DEEP Sec.4 / ChatGPT Sec.7), pairs with A-04. `EvaluationChainResolutionPolicy` (A-04) decides the route; the receipt records HOW a chain resolved at read time so Pattern C is inspectable.
```ts
interface EvaluationChainResolutionReceipt {
chain_id: string; resolved_target_ref: string; // the evaluated_target the Judge read (A-02)
resolution_basis: "single" | "pattern_c_parallel" | "policy_default";
policy_ref: string; resolved_at: ISO8601; schema_version: 1;
}
// lint: validation.chain_resolved_without_receipt
```
### A-03 -- [Common Contracts Sec. 3.7] Pattern C chain-ID lifecycle + chain registry (ACCEPT)
- **Raised by:** Claude **B3**; ChatGPT CG-S2; Grok GK-3.6 / GK-4.8.
- **Problem:** Sec. 5.18.4 / Sec. 3.7 say the upstream Evaluator populates `target_evaluation_chain_id` and the Judge copies it, but nothing specifies: (1) who generates the UUID and whether it's minted even when no Pattern C Judge attaches; (2) what happens when it doesn't resolve at the consumer; (3) retention/GC; (4) whether re-activations share an ID. The field has no lifecycle, so audit reconstruction only works when Pattern C wiring happens to be present.
- **Disposition:** **ACCEPT.** Evaluator MUST mint a fresh **ULID** at activation and emit it; the Pattern C Judge MUST read it from `evaluator_output_in.target_evaluation_chain_id` and set its own envelope to the same value; orphan envelopes keep the field unused; retention = envelope retention; add `validation.pattern_c_chain_id_mismatch`. Back it with a chain registry so chain status is first-class.
- **Fix (Common Contracts Sec. 3.7) -- `[Appx E]`:**
```ts
interface EvaluationChainRegistryRecord {
chain_id: string; chain_kind: EvaluationChainKind; task_id: string; run_id: string;
target_artifact_ref: StorageRef | null; target_artifact_version_ref: StorageRef | null;
target_scope_ref: ArtifactScopeRef | null; evaluation_snapshot_ref: StorageRef;
expected_producers: ProducerKind[]; received_result_ids: string[];
status: EvaluationChainStatus; created_at: ISO8601; completed_at?: ISO8601; superseded_by_chain_id?: string;
validation_failures: Array<"chain_id_missing" | "chain_target_mismatch" | "chain_stale_snapshot" | "chain_ambiguous" | "chain_consumer_timeout">;
schema_version: "1.0";
}
```
**A-04** -> rendered in Layer 3 core (that rendering governs).
**A-05** -> rendered in Layer 1.3 / severity reframe (that rendering governs).
### A-06 -- [V3.3.1 Sec. 0.4.1 / Common Contracts Sec. 3.1-3.2] `OutcomeEvaluationState` count & mapping skew (ACCEPT)
- **Raised by:** Claude **B2**; ChatGPT CG-S10/S11/S12.
- **Problem:** V3.3.1 enumerates **15** `OutcomeEvaluationState` values; Common Contracts Sec. 3.1 calls it a "14-value enum"; Sec. 3.2's verdict mapping covers 14 -- `evaluating` is unmapped -- and `max_iterations_reached` is referenced elsewhere but absent from the enum. The document miscounts its own enum.
- **Disposition:** **ACCEPT.** Split **runtime** states (4 -- never emitted to an envelope) from **disposition** states (12 -- emitted, each with a verdict mapping); add `max_iterations_reached`; make a single matrix the source of truth for terminal/verdict/feedback-branch/UI-label/learning-eligibility per state.
- **Fix (V3.3.1 Sec. 0.4.1 + Common Contracts Sec. 3.2) -- `[Appx B]`:**
```ts
type OutcomeEvaluationRuntimeState = "pending" | "pending_dependency" | "evaluating" | "dirty";
type OutcomeEvaluationDisposition =
| "satisfied" | "needs_revision" | "needs_information" | "needs_verification"
| "needs_human_judgment" | "unable_to_evaluate" | "blocked_by_policy" | "regressed"
| "upstream_failure" | "unrecoverable" | "superseded" | "max_iterations_reached";
type OutcomeEvaluationState = OutcomeEvaluationRuntimeState | OutcomeEvaluationDisposition;
// + OUTCOME_STATE_MATRIX (Appx B): per-state {state_class, emitted_to_envelope, terminal,
// verdict_mapping, feedback_branch, blocks_downstream_default, revisor_default_action,
// ui_label, learning_eligibility} -- this matrix is normative; Sec. 3.2 references it rather than re-listing.
```
- **Rationale:** runtime states (`pending`/`evaluating`/`dirty`/`pending_dependency`) explicitly carry `emitted_to_evaluation_envelope:false`, which is exactly why `evaluating` was never mapped -- it should never reach the verdict mapping at all.
**A-07** -> rendered in Layer 3 core (that rendering governs).
### A-08 -- [FD Sec. 6.2] Routing completeness: missing branches + full state->routing matrix (ACCEPT)
- **Raised by:** Claude **B4**; ChatGPT CG-S15/S16.
- **Problem:** `FeedbackRoutingPolicy` has `on_satisfied`/`on_needs_revision`/`on_needs_more_sources`/`on_needs_source_verification`/`on_needs_format_repair`/`on_repeated_failure` but **no** `on_indeterminate`, `on_not_applicable`, `on_unrecoverable`, `on_blocked_by_policy`, `on_upstream_failure` -- and indeterminate is not rare (5 states map to it). It also can't represent multiple simultaneous actions.
- **Disposition:** **ACCEPT.** Adopt the closed `FeedbackBranch` set; the state->branch routing is the matrix's `feedback_branch` column (A-06), so routing and state stay in lockstep. Delivery/consumption receipts come from Appendix F.
- **Fix (FD Sec. 6.2) -- `[Appx B]` branch set + `[Appx F]` receipts:**
```ts
type FeedbackBranch =
| "on_satisfied" | "on_needs_revision" | "on_needs_more_sources" | "on_needs_source_verification"
| "on_needs_human_judgment" | "on_blocked_by_policy" | "on_upstream_failure"
| "on_unrecoverable" | "on_repeated_failure" | "none";
// FeedbackRoutingPolicy keys MUST cover every FeedbackBranch; the active branch is taken
// from OUTCOME_STATE_MATRIX[state].feedback_branch. Delivery + consumption receipts: Appx F.
```
### A-09 -- [Common Contracts Sec. 4.2 / V3.3.1 Sec. 5.17.7] Phantom & misplaced schema refs + TypeOwnerRegistry (ACCEPT)
- **Raised by:** Claude **B5 + B6**; ChatGPT (ref findings).
- **Problem:** Four ownership errors. (1) V3.3.1 Sec. 5.17.7 says `ClaimSetBundle`/`ExtractedEvaluationUnit` live in Common Contracts, but Sec. 1.2 explicitly puts them **out of scope** (owner = Addenda A). (2) Common Contracts Sec. 4.2 says `ResearchNeed` lives in Core -- it's in **Source Workspace Sec. 6.2**. (3) Same Sec. 4.2 says `OutcomeRepairInstruction` lives in Core -- it's in **FD Sec. 5.2**. (4) `EvaluationAffirmation` is referenced but **defined nowhere** (phantom). A coding agent following these refs looks in the wrong document four times.
- **Disposition:** **ACCEPT.** Install a `TypeOwnerRegistry` as the single ownership source and fix the four pointers. **Decide `EvaluationAffirmation`: DELETE** it from the qualitative slice -- its real need (the positive/"what the artifact got right" counterpart) is met by the verdict-aware `OutcomeEvaluationSignal` denominator, which is **Phase-B-gated** (see E-05), not by a phantom finding type.
- **Fix (Common Contracts, new governance section) -- `[Appx A]`:**
```ts
// TypeOwnerRegistry: one entry per shared type; validation.type_owner_drift.* on mismatch.
// Canonical homes set by this registry:
// EvaluationResultEnvelope -> Common Contracts Sec. 3 (canonical)
// EvaluationFinding -> Common Contracts Sec. 4.2A v2.0 (canonical; projection: FeedbackFindingView)
// OutcomeRepairInstruction -> FD Sec. 5 (canonical) <- fixes ref (3)
// ResearchNeed -> Source Workspace Sec. 6 (canonical) <- fixes ref (2)
// ClaimSetBundle / ExtractedEvaluationUnit -> Addenda A (imported) <- fixes ref (1)
// CostEstimate -> Common Contracts Sec. 9A (canonical)
// V3.3.1 Sec. 5.17.7: cross-ref ExtractedEvaluationUnit/ClaimSetBundle -> Addenda A; keep only ArtifactScopeRef -> Common Contracts Sec. 7.
// Common Contracts Sec. 4.2: EvaluationAffirmation -> REMOVED from QualitativeSlice (see E-05).
```
- **(!)verify:** confirm `EvaluationAffirmation` is truly absent before deleting (review grep reports zero hits -- low risk).
### A-10 -- [Common Contracts Sec. 3 / Sec. 11] `EvaluationArtifactEnvelope` wrapper out-of-document; `overall_state` too B-specific (ACCEPT WITH MODIFICATIONS)
- **Raised by:** ChatGPT CG-S22, CG-AA4.
- **Problem:** the wrapper that governs envelopes is referenced from outside the set, and `overall_state` is shaped for Addenda B, so other producers (e.g. the Addenda A Experiment) can't reuse it.
- **Disposition:** **ACCEPT WITH MODIFICATIONS.** Bring `EvaluationArtifactEnvelope` into Common Contracts as the universal governance wrapper (the Addenda A coordination rows R199/R213 already adopt it on that side); generalize `overall_state` to the canonical disposition set (A-06) so every producer maps onto one vocabulary.
- **Fix:** TypeOwnerRegistry entry (`EvaluationResultEnvelope`/`EvaluationArtifactEnvelope` canonical in Common Contracts) + `overall_state: OutcomeEvaluationDisposition` (from A-06) replacing the B-specific union.
**A-11** -> rendered in Layer 2 (flip) (that rendering governs).
**A-12** -> rendered in Layer 3 core (that rendering governs).
### A-13 -- [Core Sec. 9 vs Common Sec. 5] Core duplicates the learning-signal envelope with malformed TypeScript (ACCEPT)
- **Raised by:** ChatGPT CG-S21.
- **Problem:** Core re-declares the common `EvaluationLearningSignalEnvelope`, and the duplicated TypeScript is malformed -- two sources of truth, one of them broken.
- **Disposition:** **ACCEPT.** Single canonical envelope in Common Contracts Sec. 5; Core Sec. 9 **imports** it and deletes its copy.
- **Fix:** TypeOwnerRegistry: `EvaluationLearningSignalEnvelope` canonical = Common Contracts Sec. 5; add normative line to Core Sec. 9 -- "imports `EvaluationLearningSignalEnvelope` from Common Contracts Sec. 5; MUST NOT redeclare." Remove the malformed Core block.
### A-14 -- [Common Contracts Sec. 4.2 / validation] Qualitative-slice owner map wrong; slice-required check only a warning (ACCEPT)
- **Raised by:** ChatGPT CG-S20; Grok GK-4.8.
- **Problem:** two issues -- the qualitative-slice owner map is wrong (same root as A-09), and `validation.envelope_judge_emitted_qualitative_slice` is only a **warning** even though Pattern C **requires** the slice, so a non-conforming Pattern C envelope passes validation.
- **Disposition:** **ACCEPT.** Owner map fixed by the TypeOwnerRegistry (A-09); promote the validation to an **error** when `chain_kind = pattern_c_evaluator_then_judge`.
- **Fix:**
```ts
// validation.envelope_judge_emitted_qualitative_slice:
// severity = "error" WHEN chain_kind === "pattern_c_evaluator_then_judge"
// severity = "warning" otherwise
```
### A-15 -- [Common Contracts Sec. 11.5 / cross-doc obligations] "Build-ready" overstated; command registry not mandatory; compat claim too strong (ACCEPT WITH MODIFICATIONS)
- **Raised by:** ChatGPT CG-C4/C6/C7; Grok GK-4.4; Claude **D5/D22**.
- **Problem:** the set is labeled build-ready while undefined imported types and pending cross-doc obligations remain; the command registry (Core routes/commands) is described but **not mandatory**; the Sec. 11.5 backward-compat claim overstates stability; section-number cross-refs are used where stable anchors are needed before the DOC23 R3.2 absorption.
- **Disposition:** **ACCEPT WITH MODIFICATIONS.** Gate the "build-ready" status on (a) TypeOwnerRegistry populated with no `pending_absorption` blockers, and (b) OP-A obligations closed; make the command registry **mandatory** (Appendix I); soften Sec. 11.5 to "compatible within the locked schema-of-record set, no cross-version guarantee yet"; convert section-number cross-refs to stable anchors as pre-absorption hygiene.
- **Fix:** prose obligations above + Appendix I (command registry, see Sec. D-14 next pass) + anchor-stabilization rule applied across the set before R3.2 absorption.
---
**A-16** -> rendered in Layer 3 core (that rendering governs).
### A-17 -- [FD Sec. 5.3 / V3.3.1 Sec. 9] Revisor input confused with `revision_in` (ACCEPT)
- **Raised by:** ChatGPT (BUG/HIGH).
- **Problem:** FD says the Revisor consumes `OutcomeRepairInstruction` via `revision_in`. But V3.3.1 treats `revision_in` as the port on **revision-capable modules that perform repairs**; the Revisor is the **planner** that compiles repair instructions into a plan. Wiring the planner's input as `revision_in` reverses the architecture and risks an implementer building the Revisor as a revision *target*.
- **Disposition:** **ACCEPT.** Give the Revisor planner-appropriate input ports; reserve `revision_in` for artifact-mutating modules.
- **Fix (FD Sec. 5.3 + V3.3.1 Sec. 9):**
```ts
// Revisor (planner) input ports -- NOT revision_in:
// feedback_bundle_in, repair_instruction_in, evaluation_result_in
// revision_in is reserved for revision-capable (artifact-mutating) modules that EXECUTE the compiled plan.
// validation.revisor_declares_revision_in (error).
```
### A-18 -- [FD Sec. 1.2 / Sec. 2.3] Feedback-bundle emission discipline contradicts itself (ACCEPT)
- **Raised by:** ChatGPT (BUG/HIGH).
- **Problem:** FD says both the envelope and the feedback bundle are emitted by every evaluator producer, then says deterministic scorers emit a bundle only on failure and pass cases may emit envelope-only. So a missing bundle is ambiguous: not emitted by design, lost, not applicable, or policy-filtered -- a consumer can't tell which.
- **Disposition:** **ACCEPT.** Add an emission matrix by producer-kind x verdict; require either a bundle or an explicit absence reason.
- **Fix (FD Sec. 1.2 / Sec. 2.3):**
```ts
interface FeedbackEmissionMatrixEntry {
producer_kind: "outcome_evaluator" | "judge" | "deterministic_scorer";
verdict: EvaluationVerdict;
emits_envelope: boolean;
emits_feedback_bundle: "always" | "on_failure_only" | "never";
}
type FeedbackBundleAbsentReason = "not_emitted_by_design" | "not_applicable" | "policy_filtered";
// Every result MUST carry a feedback bundle OR a feedback_bundle_absent_reason. validation.feedback_bundle_absence_unexplained (error).
```
### A-19 -- [V3.3.1 Sec. 0.4.4 / Sec. 5.7.1] `FindingState` pass-through `proposed` has no negative exit (ACCEPT) -- Claude C2
- **Raised by:** Claude C2 (BUG/HIGH).
- **Problem:** the Sec. 5.7.1 transition table gives `proposed -> active` as the only outbound transition from `proposed`. There is no `proposed -> dismissed` for a finding the Evaluator decides **not** to confirm, so unconfirmed findings sit in `proposed` forever or implementations invent an unlisted transition.
- **Disposition:** **ACCEPT.** Add the negative exit + an auto-dismiss at activation termination. (The canonical `FindingState` from A-01 already includes `dismissed`; this supplies the missing *transition*.)
- **Fix (V3.3.1 Sec. 5.7.1):**
```
add transition: proposed -> dismissed
predicate: "Evaluator did not confirm the finding before activation completion"
A finding still in `proposed` at activation termination auto-transitions to:
state = dismissed, dismissal_reason = "not_confirmed_at_termination".
```
### A-20 -- [Common Contracts Sec. 6.4 / Sec. 9.4] `unanchored_llm_judgment` acknowledgment has no field to write into (ACCEPT) -- Claude C6
- **Raised by:** Claude C6 (BUG/MEDIUM).
- **Problem:** Sec. 9.4 requires explicit user acknowledgment when a required criterion is scored by `unanchored_llm_judgment`, but no schema has a field to store that acknowledgment -- so the warning fires every time (loud) or the ack lands in an undocumented field. (Partially related to C-01's `unanchored_llm_judgment_policy`, which governs *aggregation eligibility*, not the *acknowledgment record*.)
- **Disposition:** **ACCEPT.** Add the acknowledgment fields to `Criterion`.
- **Fix (Common Contracts, `Criterion`):**
```ts
// add to Criterion:
unanchored_aggregation_acknowledged_by_user: boolean; // default false
unanchored_ack_user_ref?: string;
unanchored_ack_at?: ISO8601;
// Warning fires WHEN scoring_basis == "unanchored_llm_judgment" && required == true
// && unanchored_aggregation_acknowledged_by_user == false; silences once acknowledged.
```
### A-21 -- [FD Sec. 2.2 / Sec. 3.3] Optional fields are required by invariants (ACCEPT -- largely satisfied by A-01)
- **Raised by:** ChatGPT (BUG/MEDIUM).
- **Problem:** FD says `findings[i].based_on_artifact_version_ref` must resolve to a version in the snapshot, but the field is optional -- so validators can't enforce the invariant consistently.
- **Disposition:** **ACCEPT.** Already addressed structurally by A-01's canonical `EvaluationFinding`, which carries both `based_on_artifact_version_ref` and `based_on_artifact_version_absent_reason`. Add the enforcing validation.
- **Fix:**
```
// Satisfied by A-01 canonical EvaluationFinding. Add:
// validation.artifact_targeted_finding_missing_version_ref (error): finding_kind targets an artifact AND
// both based_on_artifact_version_ref and based_on_artifact_version_absent_reason are null.
```
### A-22 -- [Common Contracts Sec. 12] Pending consumers lack degraded-mode behavior (ACCEPT)
- **Raised by:** ChatGPT (GAP/HIGH).
- **Problem:** Common Contracts lists pending target updates to DOC8/BDSM, EC Core, DOC20, DOC72, PropA, etc. Until those land, producers can emit envelopes/signals **no consumer can interpret**, with no defined behavior.
- **Disposition:** **ACCEPT.** Each pending cross-doc obligation declares an "until-target-lands" behavior, recorded on its TypeOwnerRegistry entry (`status: pending_absorption`, A-09).
- **Fix:**
```ts
type PendingConsumerDegradedBehavior =
"persist_only" | "suppress_promotion" | "disable_ui_affordance" | "emit_validation_warning" | "block_route";
// Every TypeOwnerRegistryEntry with status "pending_absorption" carries pending_consumer_behavior: PendingConsumerDegradedBehavior.
```
- **Rationale:** also feeds A-15 -- "build-ready" can't be claimed while any pending obligation lacks a declared degraded behavior.
**A-23** -> rendered in Layer 3 core (that rendering governs).
### A-24 -- [Common Contracts Sec. 3.4 vs V3.3.1 Sec. 5.18.8] Parallel Judge+Evaluator example contradicts topology (ACCEPT) -- Claude C14
- **Raised by:** Claude C14 (BUG/LOW).
- **Problem:** Sec. 3.4's example ("Judge and Evaluator running in parallel on the same artifact version") fits no specified topology -- Pattern C has the Judge *consume* the Evaluator's output, so they can't run in parallel.
- **Disposition:** **ACCEPT.** Fix the example.
- **Fix:** replace the Sec. 3.4 example with "two Evaluator activations on the same snapshot during an Experiment, or an Evaluator and a deterministic scorer in parallel" -- or specify the genuine parallel topology. Removes a wiring trap.
### A-25 -- [Common Contracts Sec. 3.1 vs V3.3.1 Sec. 5.1] `evaluation_chain_id` naming asymmetry (ACCEPT -- fold into A-03) -- Claude D8
- **Raised by:** Claude D8 (BUG/LOW).
- **Problem:** the field is `target_evaluation_chain_id` on the envelope but referred to as `evaluation_chain_id` in V3.3.1 Sec. 5.1 prose -- same concept, two names.
- **Disposition:** **ACCEPT.** Normalize to `target_evaluation_chain_id` everywhere (fold into the A-03 chain-lifecycle work). `validation.chain_id_name_drift` (lint).
---
**A-26** -> rendered in Layer 1 (that rendering governs).
### A-27 -- [V3.3.1 Sec. 5.16 / Source Workspace] `EvaluationSnapshot` source-workspace hashes typed as artifact hashes (ACCEPT)
- **Raised by:** ChatGPT Audit Addendum (BUG/HIGH).
- **Problem:** `EvaluationSnapshot.source_workspace_head_hashes` is typed `Record<ArtifactRef, string>`, but the snapshotted values are Source Workspace heads, source records, source sets, research-need queues, and verification records -- **not** artifact refs. The invalid type model breaks audit, revalidation, live-edit checks (B-03), and source-freshness detection. (Distinct from B-03, which is the *precondition*; this is the *type* that precondition compares.)
- **Disposition:** **ACCEPT.** Replace with a workspace-native snapshot-hash structure.
- **Fix (`[Appx, addendum]`):**
```ts
interface SourceWorkspaceSnapshotHashSet {
source_workspace_ref: string;
source_workspace_head_hash: string;
source_record_hashes: Record<SourceRecordRef, string>;
source_set_hashes?: Record<SourceSetRef, string>;
research_need_queue_hash?: string;
verification_record_hashes?: Record<string, string>;
freshness_record_hashes?: Record<string, string>;
run_guidance_hashes?: Record<string, string>;
schema_version: "1.0";
}
// EvaluationSnapshot: REMOVE source_workspace_head_hashes: Record<ArtifactRef,string>;
// ADD source_workspace_state_ref: StorageRef + source_workspace_snapshot_hashes: SourceWorkspaceSnapshotHashSet[].
```
- **Rationale:** this is the concrete `SnapshotHash` type that B-03's `WorkspaceWritePrecondition.expected_snapshot_hash` and D-12 retention compare against -- without it, "compare the snapshot hash" has no well-typed object.
---
---
# Sec. B. Runtime / concurrency / distributed-systems
### R.2 -- Sec.B rows
### B-01 -- [V3.3.1 Sec. 11 / Core Sec. runtime] No unified RevisionExecutionLifecycle / dispatcher<->receipts state machine (ACCEPT)
- **Raised by:** Grok GK-4.1 / GK-5.6; ChatGPT (runtime).
- **Problem:** the execution lifecycle (plan compiled -> dispatched -> step receipts -> revalidation -> terminal) is specified in pieces across Sec. 11.x with no single owning state machine. A coding agent assembles the order from prose and the dispatcher<->receipt handshake is implicit.
- **Disposition:** **ACCEPT.** Define one normative `RevisionExecutionLifecycle` state machine in V3.3.1 Sec. 11 (states + legal transitions + which receipt drives each transition), anchored by the Appendix M primitives (`PendingDependencyInfo`, `HardCallPendingPolicy`) and the cancel/known-good types below.
- **Fix:** add a Sec. 11.0 lifecycle FSM listing states `{plan_compiled, dispatched, step_running, step_receipt_received, revalidating, regression_detected, terminal_satisfied, terminal_failed, cancelled, hard_call_pending}` and the receipt that triggers each edge; every other Sec. 11.x rule references a state in this FSM rather than re-describing flow.
**B-02** -> rendered in Layer 3 core (that rendering governs).
### B-03 -- [Source Workspace Sec. x / V3.3.1 Sec. 11.20] Concurrent-write safety on the Source Workspace (ACCEPT) (!)verify
- **Raised by:** Gemini **F-01** (#1 build blocker) + D-05; ChatGPT CG-SW9; relates to Claude B9.
- **Problem:** a `direct_fix` mutation can race a live UI edit (or a parallel module write) to the same workspace record with no precondition check, silently corrupting it; no transaction boundaries or read/write locks are specified.
- **Disposition:** **ACCEPT.** Mandate a snapshot-hash precondition on every in-place workspace write; on mismatch, abort and re-evaluate rather than overwrite. Add explicit lock semantics.
- **Fix (Source Workspace write path):**
```ts
interface WorkspaceWritePrecondition {
target_record_ref: StorageRef;
expected_snapshot_hash: string; // hash read at plan time
on_mismatch: "abort_and_reevaluate" | "queue_behind_lock" | "fail_outcome"; // default abort_and_reevaluate
lock_mode: "optimistic_snapshot_hash" | "pessimistic_write_lock"; // default optimistic
}
// direct_fix MUST carry a WorkspaceWritePrecondition; dispatcher computes the real hash at execution
// and compares to expected_snapshot_hash before applying. validation.workspace_write_without_precondition (error).
```
- **(!)verify:** confirm the current `direct_fix` write path and whether any precondition exists (Gemini asserts none; this is the single item Gemini calls catastrophic if missed -- worth confirming before patch).
**Audit fold-in (R0.4) -- Source Workspace write-operation API** (DEEP Sec.5/Sec.7). The CH-05 fixture asserts the precondition; this renders the operation contract it checks.
```ts
interface SourceWorkspaceOperation {
operation_id: string; workspace_ref: string; op_kind: "append" | "update" | "supersede" | "delete";
target_artifact_ref?: string; expected_precondition_hash?: string; // optimistic concurrency (rolling hash, B-04)
lock_lease?: { lease_id: string; holder_ref: string; expires_at: ISO8601 };
governance: GovernanceEnvelope; schema_version: 1;
}
interface SourceWorkspaceOperationReceipt {
operation_id: string; result: "applied" | "precondition_failed" | "lease_lost" | "rejected";
resulting_artifact_ref?: string; applied_at: ISO8601;
}
// lints: validation.workspace_mutation_without_precondition ; validation.workspace_operation_without_receipt ; validation.workspace_lease_expired_but_write_applied
```
### B-04 -- [V3.3.1 Sec. 11.20.2 / Sec. 11.22] Rolling hash unsafe under DAG/parallel execution (ACCEPT)
- **Raised by:** ChatGPT CG-R2 / CG-R3 / CG-AA10; Claude **B9**.
- **Problem:** Sec. 11.22 allows up to `max_parallel_steps_per_plan: 4`, but Sec. 11.20.2 Rolling-Hash Mode B requires "step N+1 validates against predicted hash from step N." Two parallel steps mutating the same artifact can't both validate against one base -- the chain is nondeterministic. Sec. 11.20.2 forbids concurrent *plans* on the artifact but not concurrent *steps* within one plan. (Distinct from C-03, which is the LLM-predicted-hash problem.)
- **Disposition:** **ACCEPT.** Rolling-hash Mode B requires sequential execution across all steps mutating the same artifact; parallelism allowed only between steps on disjoint artifacts.
- **Fix (V3.3.1 Sec. 11.20.2 + Sec. 11.22):**
```
Sec. 11.20.2: "Rolling-hash Mode B requires sequential step execution across all steps that mutate the
same artifact. Parallelism within a plan is permitted only between steps targeting disjoint artifacts."
Sec. 11.22: parallel batches automatically degrade to sequential when any step is rolling-hash Mode B on a shared artifact.
validation.rolling_hash_parallel_steps_same_artifact (error)
```
### B-05 -- [V3.3.1 Sec. 11.21 / Sec. 11.21.2 / Sec. 5.14.1] Revalidation cascade: no convergence bound + duplicated upstream-failure rule (ACCEPT)
- **Raised by:** Claude **B7** (no convergence) + **B8** (duplicated rule).
- **Problem (B7):** Sec. 11.21 Phase 4 re-triggers the Revisor on regression with no bound on cascade depth; two outcomes with bidirectional `OutcomeDependencySpec.invalidated_by_outcomes` can ping-pong revisions forever -- per-outcome budget never trips because each cycle burns a *different* outcome's budget; `per_plan_max_replans` is plan-level, not cascade-level. **Problem (B8):** the `upstream_failure_cascade` rule is stated in both Sec. 5.14.1 and Sec. 11.21.2 (invites drift) and neither handles the race where an outcome enters `pending_dependency` *after* the cascade fired.
- **Disposition:** **ACCEPT.** Add a cascade-depth bound + cycle detection; consolidate the duplicated rule into one section with a state-entry guard for the race.
- **Fix (V3.3.1 Sec. 6.14 RevisorConfig + Sec. 11.21.2 + Sec. 22):**
```ts
// RevisorConfig:
max_revalidation_cascade_depth: number; // default 5, measured from the originating mutation receipt
// Loop Controller tracks cascade chains; a regression re-entering an outcome already in the chain -> abort:
// validation.revalidation_cascade_loop (new, Sec. 22)
// HardRevisionCallKind += "revalidation_cycle" // surface tie-break to the user
// Sec. 11.21.2 becomes the single home of upstream_failure_cascade; Sec. 5.14.1 links to it (no restating).
// State-entry guard: an outcome transitioning to pending_dependency AFTER a cascade fired on its
// upstream module is auto-evaluated against the upstream_failure set at state entry (not a new cascade pass).
```
**Audit fold-in (R0.4) -- revalidation cascade** (DEEP Sec.5), shared by B-05/B-06/B-20. CH-06 asserts convergence; this renders the cascade run + the dependency-graph policy that bounds it. *(Row-merge with B-06/B-20 remains declined -- see Confirmed divergences; rows stay distinct, the cascade mechanism is shared.)*
```ts
interface OutcomeDependencyGraphPolicy {
edges: { from_outcome_id: string; to_outcome_id: string }[];
cycle_handling: "reject_at_build" | "break_with_receipt"; // dag_acyclic (B-25 verified)
max_cascade_depth: number; // convergence bound
}
interface RevalidationCascadeRun {
cascade_id: string; trigger_outcome_id: string; revalidated_outcome_ids: string[];
depth_reached: number; converged: boolean; schema_version: 1;
}
// lints: validation.revalidation_cascade_unbounded ; validation.duplicate_upstream_failure_revalidation // success-condition-5 race (B-20)
```
### B-06 -- [V3.3.1 Sec. dependency / Common Contracts] Outcome-dependency cycles + `pending_dependency` deadlock (ACCEPT)
- **Raised by:** Claude C4; Gemini **D-03**.
- **Problem:** dependency direction is undefined for cycles, and an upstream `could_not_fix`/halt does not instantly cascade to downstream `pending_dependency` outcomes, so the graph hangs waiting on an artifact that will never arrive.
- **Disposition:** **ACCEPT.** Define cycle detection at dependency-declaration time (reject or break with a Hard Call) and an instant downstream cascade on terminal upstream failure.
- **Fix:** dependency graph rejects declared cycles (`validation.outcome_dependency_cycle`); on upstream terminal-failure, all transitively dependent `pending_dependency` outcomes immediately transition to `upstream_failure` (per the A-06 matrix), rather than waiting on `wait_timeout`.
### B-07 -- [V3.3.1 Sec. 11.22] Parallel sibling outputs orphaned after batch failure; parallelism not first-class (ACCEPT)
- **Raised by:** ChatGPT CG-R6 / CG-R7 / CG-AA11.
- **Problem:** when one step in a parallel batch fails, the sibling steps' completed outputs have no defined finalization/disposition (orphaned candidates), and the parallel-batch configuration isn't a first-class, inspectable object.
- **Disposition:** **ACCEPT.** Add a `ParallelBatchFinalizationReceipt` recording per-sibling disposition on batch failure; surface the parallelism config.
- **Fix:**
```ts
interface ParallelBatchFinalizationReceipt {
batch_id: string; plan_id: string; run_id: string;
sibling_results: Array<{ step_id: string; status: "completed" | "failed" | "cancelled" | "orphaned";
candidate_ref?: StorageRef; disposition: "retained" | "discarded" | "held_for_review"; }>;
batch_outcome: "all_completed" | "partial_failure" | "aborted";
created_at: ISO8601; schema_version: "1.0";
}
```
**B-08** -> rendered in Layer 3 core (that rendering governs).
### B-09 -- [V3.3.1 Sec. candidate] Candidate artifacts can't model external side effects; lifecycle mixed with head-pointer (ACCEPT)
- **Raised by:** ChatGPT CG-R13 / CG-R14.
- **Problem:** `CandidateArtifactVersion` conflates the version's lifecycle state with the "current head" pointer, and there's no way to model a candidate whose application has an external side effect (the thing that can't be branched -- see Sec. F-01).
- **Disposition:** **ACCEPT.** Separate candidate lifecycle from the head pointer; route side-effecting candidates through `SideEffectIntentCandidate` (Appendix N).
- **Fix (V3.3.1 candidate model + Appendix N):**
```ts
// Split: CandidateArtifactVersion.lifecycle_state (candidate|accepted|rejected|superseded|reverted)
// ArtifactHead.current_version_ref (separate projection; one head per artifact)
// Side-effecting application uses SideEffectIntentCandidate [Appx N]:
// { side_effect_class, dry_run_payload_ref, approval_status, execution_policy_ref,
// state: draft|approved|executed|cancelled|blocked, execution_receipt_ref }
```
### B-10 -- [V3.3.1 Sec. 11 / FD] `TaskCancelProtocol` + Hard Call blocking scope undefined (ACCEPT)
- **Raised by:** ChatGPT CG-R15 / CG-R16.
- **Problem:** there is no clean mid-run cancel protocol, and a pending blocking Hard Call doesn't say what it blocks (the step, the outcome, the artifact, the whole run, or just side effects).
- **Disposition:** **ACCEPT.** Adopt Appendix M's `TaskCancelProtocol` and `HardCallPendingPolicy`/`HardCallBlockingScope` verbatim.
- **Fix -- `[Appx M]`:**
```ts
type HardCallBlockingScope = "entire_run" | "segment" | "artifact" | "outcome" | "module" | "side_effect_only";
interface HardCallPendingPolicy { hard_call_id: string; blocking_scope: HardCallBlockingScope;
blocked_refs: string[]; allowed_to_continue_refs: string[];
context_visible_to_continuing_modules: "none" | "hard_call_pending_summary" | "full_context_redacted";
on_defer: "continue_with_warning" | "pause_scope" | "abort_scope";
on_timeout: "escalate" | "abort" | "continue_with_warning"; timeout_ms?: number; schema_version: "1.0"; }
interface TaskCancelProtocol { cancel_request_id: string; task_id: string; run_id: string; requested_by_ref: string;
cancel_scope: "entire_run" | "segment" | "module_activation" | "revision_plan" | "side_effect_intent";
target_refs: string[]; in_flight_handling: "request_graceful_stop" | "preempt_immediately" | "finish_current_step_then_stop";
side_effect_policy: "do_not_cancel_executed_side_effects" | "cancel_unexecuted_intents" | "create_corrective_artifact";
candidate_disposition: "discard" | "retain_for_manual_review" | "orphan_until_reconciled";
source_workspace_disposition: "retain_records" | "mark_records_cancelled" | "rollback_if_uncommitted";
learning_signal_policy: "suppress_success_signals" | "emit_cancel_diagnostic_only" | "emit_full_signals_with_cancel_flag";
user_receipt_ref?: StorageRef; created_at: ISO8601; schema_version: "1.0"; }
```
- **Rationale:** `side_effect_policy: do_not_cancel_executed_side_effects` is the cancel-side counterpart of the Sec. F-01 "side effects can't branch" principle -- cancel never pretends an executed effect is undone.
### B-11 -- [V3.3.1 Sec. skip / FD] `TaskSkipProtocol` + skip receipts missing (ACCEPT)
- **Raised by:** ChatGPT CG-AA5.
- **Problem:** an outcome/step can be skipped, but there's no protocol or receipt recording who skipped it, why, and the downstream effect.
- **Disposition:** **ACCEPT.** Add a skip protocol mirroring cancel, with a receipt.
- **Fix:**
```ts
interface TaskSkipReceipt { skip_id: string; target_ref: string; skipped_by_ref: string;
reason: "not_applicable" | "user_directed" | "dependency_unavailable" | "policy_blocked";
downstream_effect: "none" | "marks_dependents_not_applicable" | "requires_human_ack";
created_at: ISO8601; schema_version: "1.0"; }
```
### B-12 -- [Common Contracts Sec. policy / V3.3.1] Policy freshness checks a field `PolicyDecisionRef` lacks (ACCEPT)
- **Raised by:** ChatGPT CG-R8.
- **Problem:** freshness logic expects a `superseded_by_decision_id` (and related staleness fields), but `PolicyDecisionRef` has no such field, so freshness can't actually be evaluated.
- **Disposition:** **ACCEPT.** Add the freshness fields, or make the EC policy record the authoritative freshness source.
- **Fix:**
```ts
// add to PolicyDecisionRef (or PolicyEvaluationRef):
issued_at: ISO8601; subject_hash: string; decision_scope_hash: string;
superseded_by_decision_id?: string; policy_engine_version: string;
// alt: mark EC's policy record as the required freshness source-of-truth and have the ref point to it.
```
### B-13 -- [Source Workspace Sec. 6 / Common Contracts] `ResearchNeed` concurrency -- needs a lease (ACCEPT)
- **Raised by:** Claude **D15**.
- **Problem:** two modules can independently pick up and satisfy the same `ResearchNeed` (duplicate work, conflicting results), because there's no lease/claim with idempotency.
- **Disposition:** **ACCEPT.** Add a `ResearchNeedLease` so a need is claimed atomically.
- **Fix:**
```ts
interface ResearchNeedLease { need_id: string; leased_by_module_id: string; lease_token: string;
acquired_at: ISO8601; expires_at: ISO8601;
on_expiry: "release_for_reclaim" | "escalate" | "mark_abandoned";
satisfied_by_result_ref?: StorageRef; schema_version: "1.0"; }
// acquisition is atomic compare-and-set on need status; second acquirer gets need_already_leased.
```
### B-14 -- [Task Forum Sec. x] Forum deadlock breaker missing (ACCEPT)
- **Raised by:** Gemini **F-02**; Claude D3 / D20.
- **Problem:** the Task Forum has no tie-break, timeout, or max-rounds, so deliberation can hang with no circuit breaker.
- **Disposition:** **ACCEPT.** Forum runs as an FSM with a deliberation tick cap and a consensus threshold; on expiry it escalates to a Hard Call.
- **Fix:**
```ts
interface ForumDeliberationPolicy { room_id: string; max_deliberation_ticks: number; consensus_threshold_pct: number;
on_no_consensus: "forum_deadlock_hard_call" | "task_agent_decides" | "prefer_safest_proposal";
schema_version: "1.0"; }
// reaching max_deliberation_ticks without consensus_threshold_pct -> state forum_deadlock -> HardCall escalation.
```
**B-15** -> rendered in Layer 2 (flip) (that rendering governs).
### B-16 -- [V3.3.1 Sec. 11.9] Concurrency tie-breaker relies on wall-clock `created_at` (ACCEPT-AS-FIX)
- **Raised by:** Gemini **BUG-04**.
- **Problem:** Rule 4 of `concurrency_tie_breaker` is `RevisionPlan.created_at` ascending. In a local multi-threaded runtime, millisecond timestamps are subject to event-loop clock skew; granting a workspace write-lock on timestamp creates a race and can let a 45-second plan win the lock over a 1-second fix.
- **Disposition:** **ACCEPT-AS-FIX** (Gemini's patch verbatim) -- order by risk reduction and lock-release speed, not timestamp.
- **Fix (V3.3.1 Sec. 11.9) -- `[Gemini BUG-04]`:**
```
RULE concurrency_tie_breaker (UPDATED):
1. OutcomeDependencySpec.required_for_overall_pass = true > false
2. EvaluationOutcomeDefinition.is_high_stakes = true > false
3. EvaluationOutcomeDefinition.priority ascending (lower value = wins)
4. RevisionPlan.plan_risk_score descending (safer plans acquire lock first -- prevent catastrophic collisions)
5. RevisionCostEstimate.total_tokens ascending (smaller/faster plans execute and release the lock quicker)
```
**Audit fold-in (R0.4) -- plan-selection tiebreaker** (DEEP Sec.5 / ChatGPT Sec.8). *Correction:* `tiebreaker_epsilon` was removed from the L1.5 **metric** layer (correct -- mis-anchored there); the reviewers' actual proposal is a **plan-selection** tiebreaker on `plan_risk_score`, rendered here. Wall-clock is forbidden as a tiebreaker (non-total-order under clock skew).
```ts
interface PlanSelectionTiebreakPolicy {
primary_key: "plan_risk_score"; epsilon: number; // band within which scores are treated as tied
fallthrough: ("fewest_steps" | "fewest_irreversible_effects" | "lowest_plan_id_hash")[];
schema_version: 1;
}
// lints: validation.concurrency_tie_breaker_uses_wall_clock ; validation.concurrency_tie_breaker_missing_final_tiebreaker ; validation.concurrency_tie_breaker_not_total_order
```
### B-17 -- [FD Sec. 8.4 / Sec. 10.3] `instruction_in` overload leaks the DOC23/DOC15/DOC24 boundary (ACCEPT)
- **Raised by:** Claude **B11**.
- **Problem:** FD Sec. 8.4 carries both free-form instructions and typed `OutcomeRepairInstruction` payloads over the same general `instruction_in` port, with no discriminator. Receiving modules must runtime-guess the payload type; DOC15/CIL prompt assembly can't know what to render. The typed ports (`repair_instruction_in`, etc.) are marked "ergonomics, not required for V1," so every V1 implementation goes through the overload and solves discrimination differently.
- **Disposition:** **ACCEPT.** Don't ship the overload. Either elevate the typed ports to required for V1 (preferred) or add a payload discriminator.
- **Fix (FD Sec. 8.4 + DOC23 R3.1 port registry Sec. 10.3):**
```ts
// Preferred: make these REQUIRED V1 ports (register in DOC23 R3.1 Sec. 10.3):
// feedback_in, repair_instruction_in, run_guidance_in, source_need_in
// Fallback if kept on instruction_in: payload union carries a discriminator DOC15/CIL dispatches on:
type FeedbackPayloadKind = "free_form_instruction" | "outcome_repair_instruction" | "run_guidance" | "research_need";
// validation.instruction_in_untyped_feedback_payload (error) if a typed payload rides instruction_in without a discriminator.
```
---
### B-18 -- [V3.3.1 Sec. 6.5.2 / Sec. 7.9] `HardRevisionCall.options` may be empty though spec requires bounded options (ACCEPT) -- Claude C3
- **Raised by:** Claude C3 (BUG/HIGH).
- **Problem:** Sec. 6.5.1 says detection produces a Hard Call "with bounded `HumanDecisionOption[]`," but Sec. 7.9.1's schema has no non-empty constraint. Empty `options[]` makes the Sec. 21.4 UI non-functional (no buttons); the user can't resolve and the Dispatcher stays in `waiting_hard_call` indefinitely. The `default_if_no_response` trigger ("no response") is itself unspecified.
- **Disposition:** **ACCEPT.** Constrain options to >=2, provide a default pair, and specify the no-response timeout.
- **Fix (V3.3.1 Sec. 7.9.1):**
```
options: HumanDecisionOption[] // MIN_LENGTH = 2
// When the Compiler cannot enumerate substantive options, default to:
// ["continue_with_compiler_proposal", "pause_for_my_input"]
// validation.hard_call_options_empty (error).
// Specify the timeout that triggers default_if_no_response (e.g., HardCallPendingPolicy.timeout_ms, B-10).
```
### B-19 -- [V3.3.1 Sec. 11.6 / Sec. 0.4.7] `RevisionOperationKind="hard_call_resolved"` has no producer (ACCEPT) -- Claude C10
- **Raised by:** Claude C10 (BUG/MEDIUM).
- **Problem:** `hard_call_resolved` is a valid `RevisionOperationKind` and `HardCallResolution` is persisted, but no section says which actor emits the operation receipt. Operation receipts feed `RepairCycleSignal`; ambiguous actor -> missing or duplicated receipt, breaking the `hard_call_resolved -> revision_operation_receipt_ref` chain in `RevisorActionRecord`.
- **Disposition:** **ACCEPT.** Name the producer.
- **Fix (V3.3.1 Sec. 7.9.4, new):**
```
On recording a HardCallResolution, the Dispatcher emits a RevisionOperationReceipt with
operation_kind = "hard_call_resolved" and hard_call_ref -> the resolved Hard Call.
receipt.actor_ref = Dispatcher runtime identity;
resolution.resolved_by = UserRef (recorded separately).
```
### B-20 -- [V3.3.1 Sec. 6.7.2 / Sec. 11.21] Success-condition 5 races the cascade (ACCEPT) -- Claude D2
- **Raised by:** Claude D2 (RISK/MEDIUM).
- **Problem:** success-condition 5 ("cascaded dependent outcomes are re-evaluated") is checked at revision-cycle completion, but the Sec. 11.21 cascade can still be firing -- so a revision can be marked successful before its own regression cascade settles.
- **Disposition:** **ACCEPT.** Gate condition 5 on cascade quiescence.
- **Fix:** the Loop Controller may evaluate success-condition 5 only when no revalidation is pending in the cascade chain (tracked via B-05's `max_revalidation_cascade_depth` chain state); `validation.success_marked_before_cascade_quiescent` (error).
### B-21 -- [Core Sec. 9.0.6] Signal emission ordering vs receipt persistence undefined (ACCEPT) -- Claude D7
- **Raised by:** Claude D7 (RISK/MEDIUM).
- **Problem:** Sec. 9.0.6 shows signals emitted then passing the EC policy gate, but doesn't order signal emission against durable persistence of the receipts those signals reference -- a consumer can receive a signal pointing at a not-yet-written receipt.
- **Disposition:** **ACCEPT.** Enforce emit-after-persist.
- **Fix (Core Sec. 9.0.6):** a learning signal MUST NOT be emitted until the receipts it references are durably written (read-your-writes guarantee at the EC policy gate); `validation.signal_references_unpersisted_receipt` (error).
### B-22 -- [V3.3.1 Sec. 5.18 / Sec. 11.15] Pattern C doubles per-turn latency/cost with no budget (ACCEPT) -- Claude D17
- **Raised by:** Claude D17 (RISK/MEDIUM).
- **Problem:** Pattern C wires a Judge downstream of every standalone Evaluator; in an iterative revision loop this doubles evaluation latency and cost per turn, but no budget governs the Pattern C Judge separately from the Evaluator -- a long loop silently doubles cost.
- **Disposition:** **ACCEPT.** Add a Pattern C invocation budget / cadence.
- **Fix:**
```ts
interface PatternCInvocationBudget {
max_judge_invocations_per_run?: number;
cadence: "every_turn" | "every_n_turns" | "on_terminal_only";
n_turns?: number; // required when cadence = "every_n_turns"
}
// governs the Pattern C Judge independently of the upstream Evaluator.
```
### B-23 -- [FD Sec. 6.3] Multiple delivery branches fire simultaneously with no idempotency control (ACCEPT) -- Claude D24
- **Raised by:** Claude D24 (RISK/MEDIUM).
- **Problem:** `FeedbackRoutingPolicy` branches aren't stated to be mutually exclusive; one result can match several (e.g., `on_needs_revision` and `on_needs_more_sources`), firing multiple deliveries with no idempotency key or cost guard -- duplicate or conflicting deliveries. (Complements A-08 routing completeness with the *exclusivity/idempotency* rule.)
- **Disposition:** **ACCEPT.** State branch exclusivity, or an explicit multi-fire policy with idempotency keys.
- **Fix (FD Sec. 6.3):**
```ts
type RoutingFirePolicy = "first_match_only" | "all_matching_with_idempotency";
// when "all_matching_with_idempotency", each delivery carries
// delivery_idempotency_key = hash(run_id + result_id + branch); duplicates suppressed.
```
---
**B-24** -> rendered in Layer 3 core (that rendering governs).
**B-25** -> rendered in Layer 2 (flip) (that rendering governs).
**B-26** -> rendered in Layer 3 core (that rendering governs).
**B-27** -> rendered in Layer 3 core (that rendering governs).
### R.3 -- Sec.C rows
**C-01** -> rendered in Layer 3 C-cluster (that rendering governs).
**C-02** -> rendered in Layer 3 C-cluster (that rendering governs).
**C-03** -> rendered in Layer 3 C-cluster (that rendering governs).
**C-04** -> rendered in Layer 3 C-cluster (that rendering governs).
**C-05** -> rendered in Layer 3 C-cluster (that rendering governs).
**C-06** -> rendered in Layer 3 C-cluster (that rendering governs).
**C-07** -> rendered in Layer 3 C-cluster (that rendering governs).
**Audit fold-in (R0.4) -- richer cost model** (DEEP Sec.6 / ChatGPT Sec.9). Supersedes scalar-only `CostEstimate`, which becomes a projection of `CostVector`.
```ts
type CostDimension = "tokens_in" | "tokens_out" | "wall_clock_ms" | "tool_invocations" | "usd_estimate" | "sub_agent_activations";
interface CostVector { dimensions: Record<CostDimension, number>; schema_version: 1; }
interface CostCorrelationPolicy { // how per-step CostVectors combine into a run total + quantile (C-05) summing
combine: "sum" | "max" | "p50" | "p90";
quantile_sum_treatment: "assume_independent" | "assume_perfectly_correlated" | "explicit_copula_ref";
correlation_ref?: string;
}
// lints: validation.cost_vector_unit_mismatch ; validation.cost_quantile_sum_without_correlation_policy
```
**C-08** -> rendered in Layer 3 C-cluster (that rendering governs).
**C-09** -> rendered in Layer 3 C-cluster (that rendering governs).
**C-10** -> rendered in Layer 3 C-cluster (that rendering governs).
### R.4 -- Sec.D rows
**D-01** -> rendered in Layer 3 core (that rendering governs).
### D-02 -- [Source Workspace Sec. 2 / Governance] Workspace taint aggregation undefined (ACCEPT)
- **Raised by:** ChatGPT (GAP/HIGH).
- **Problem:** a workspace holds many `SourceRecord`s of differing taint, but no rule says what the workspace's aggregate taint is -- so downstream consumers can't reason about the set.
- **Disposition:** **ACCEPT.** Adopt `TaintAggregationPolicy` (Appx G): `workspace_taint = max_taint` over records; any privileged record marks the workspace privileged; matter-scope rule explicit. (Same policy object as D-01.)
### D-03 -- [Source Workspace Sec. 0.3 / V3.3.1 Sec. 12] `TaskSourceWorkspace` vs `SourceWorkspace` identity split (ACCEPT)
- **Raised by:** ChatGPT (BUG/HIGH).
- **Problem:** the task-scoped `TaskSourceWorkspace` and V3.3's `SourceWorkspace` are described as if distinct, with no statement of whether they're one type or two -- a coding agent builds two stores.
- **Disposition:** **ACCEPT.** One canonical workspace identity in the TypeOwnerRegistry (A-09); the task-scoped form is a view/parameterization, not a separate type. `validation.dual_source_workspace_identity` (error).
**D-04** -> rendered in Layer 2 (flip) (that rendering governs).
### D-05 -- [Source Workspace Sec. 6.2 / Sec. 7.4] `ResearchNeed` scoping, ref types, and `human_needed` exit (ACCEPT) -- incl. Claude D6
- **Raised by:** ChatGPT (`run_id` conflicts with task-scoped workspaces; wrong ref types; routed to wrong port); Claude **D6** (`human_needed` no exit).
- **Problem:** `ResearchNeed.run_id` conflicts with task-scoped workspaces; source/target refs use the wrong reference types; and the `human_needed` status has no exit (who resolves it, what transitions out, what happens at run end).
- **Disposition:** **ACCEPT.** Adopt the canonical `ResearchNeed` (Appx G) with an explicit `need_scope`, corrected refs, and a `human_needed` exit + run-end default.
- **Fix -- `[Appx G]`:**
```ts
// ResearchNeed adds: need_scope: "run" | "task" | "workspace"; status includes "human_needed" with exits ->
// "answered" | "unresolved" | "cancelled"; target uses ArtifactScopeRef|ClaimRef (not raw string).
// Run-end default: any ResearchNeed still "human_needed"/"open" at run end -> "unresolved" with a carry note.
// (Lease for concurrency = ResearchNeedLease, B-13.)
```
### D-06 -- [Source Workspace Sec. 4 / Sec. 5 / Sec. 6.3] Evidence anchors, payload registry, weak tool-receipts, extractor-vs-source confusion (ACCEPT)
- **Raised by:** ChatGPT (evidence anchors; `domain_payload` registry/versioning; tool-receipt-as-research too weak; missing-extractor mis-reported as missing-source).
- **Problem:** source records can't anchor evidence to specific claims; `domain_payload` is unversioned/unregistered; a tool receipt is accepted as "material research" with too little structure; and a missing claim-extractor is reported as a missing source.
- **Disposition:** **ACCEPT.** Add first-class evidence anchors + a payload registry; distinguish extractor-missing from source-missing.
- **Fix -- `[Appx G]`:**
```ts
// SourceEvidenceAnchor { anchor_kind: page|quote|section|timestamp|row|url_fragment|paragraph|line_range;
// supports_claim_refs[]; support_strength: direct|indirect|contextual|contradicts; extracted_text_ref }
// DomainPayloadRegistryRef { domain_payload_kind; domain_payload_schema_ref; domain_payload_version }
// New status distinguishing claim_extractor_unavailable from source_unavailable.
```
**Audit fold-in (R0.4) -- claim-support derivation** (DEEP Sec.7 / ChatGPT Sec.10). `SourceEvidenceAnchor` (above) ties a claim to its source span; the derivation policy/receipt records HOW a support status was computed, so "supported" is never asserted bare.
```ts
interface ClaimSupportDerivationPolicy {
method: "exact_span" | "paraphrase_match" | "model_judgment";
min_anchor_count: number; model_judgment_requires_anchor: boolean;
}
interface ClaimSupportDerivationReceipt {
claim_ref: string; anchors: string[]; // -> SourceEvidenceAnchor[]
support_status: "supported" | "partially_supported" | "unsupported" | "not_checked";
derived_by: ClaimSupportDerivationPolicy["method"]; schema_version: 1;
}
// lints: validation.claim_support_status_without_derivation ; validation.evidence_anchor_without_claim_ref ; validation.extractor_missing_reported_as_source_missing
```
### D-07 -- [Task Forum Sec. 5.2 / Sec. 5.3] Forum posts: visibility, supersession, governance envelopes (ACCEPT)
- **Raised by:** ChatGPT (`selected_modules` visibility broken; supersession described-not-modeled; posts need governance envelopes); Grok GK-4.7/5.4.
- **Problem:** `selected_modules` visibility can't actually represent the selected modules; supersession is prose-only (no fields); and posts carry no governance (data_class/taint/policy/privilege).
- **Disposition:** **ACCEPT.** Adopt the hardened `TaskRunBoardPost` (Appx H).
- **Fix -- `[Appx H]`:**
```ts
// TaskRunBoardPost adds: visibility_target_refs: VisibilityTargetRef[] (fixes selected_modules);
// lifecycle_state: "active"|"superseded"|"retracted" + supersedes_post_ids[] + superseded_by_post_id + supersession_reason;
// governance: RunBoardGovernanceEnvelope { data_class, taint_class, policy_decision_refs, sanitization_required,
// governance_class, privileged, matter_id }.
```
### D-08 -- [Task Forum Sec. 6.3 / Sec. 6.4] Context-packet ownership, request/receipt, omission manifest, audience-enum divergence, total budget (ACCEPT) -- incl. Claude D18
- **Raised by:** ChatGPT (packet needs request/receipt/freshness/omission manifest; digest/packet audience enums diverge); Grok GK-4.3; Claude **D18** (token budget fragmented across packets).
- **Problem:** `TaskRunContextPacket` overlaps DOC24's packet ownership, lacks a request/receipt/freshness/omission manifest, and its audience enum diverges from the digest's. Separately (D18), token budgets are declared per-packet with no authority checking the **sum** a module receives.
- **Disposition:** **ACCEPT.** Adopt the request/receipt/omission types (Appx H) and a per-activation total-context budget enforced by the assembler.
- **Fix -- `[Appx H]`:**
```ts
// TaskRunContextPacketRequest { audience: TaskContextAudience, requested_max_tokens, required/optional_item_refs,
// staleness_policy } -> TaskRunContextPacketAssemblyReceipt { packet_content_hash, actual_token_count,
// omitted_items: OmittedPacketItem[], valid_until_event_seq, invalidated_by_event_kinds }.
// Single TaskContextAudience enum shared by digest + packet.
// D18: PerActivationContextBudget -- the assembler MUST sum context packet + board digest + feedback bundle and
// enforce one cap; validation.per_activation_context_budget_exceeded (error).
```
### D-09 -- [Task Forum Sec. 1 / Sec. 5] Passive board auto-publishes every event -- privacy/volume controls (ACCEPT) -- Claude B12
- **Raised by:** Claude **B12**; ChatGPT (RISK/HIGH).
- **Problem:** the passive board "auto-publishes every event," which is unbounded and leaks content with no privacy/volume gate.
- **Disposition:** **ACCEPT.** Every post carries the `RunBoardGovernanceEnvelope` (D-07) and the board applies a publication policy (volume cap + class filter).
- **Fix:** add `RunBoardPublicationPolicy { auto_publish_post_kinds[], max_posts_per_run, suppress_classes: ("privileged"|"local_only")[], rate_limit_per_min }`; posts failing the policy are withheld with a receipt. Pairs with D-12 retention.
### D-10 -- [Task Forum Sec. 7.2 / Sec. 3.2 / Sec. 4.6 / Sec. 3.1] ModuleAssistanceRequest schema, participant/moderator model, payload schemas, moderator-failure (ACCEPT) -- incl. Claude D3
- **Raised by:** ChatGPT (request lacks endpoint/lease/timeout/answer schema; participant policy only models modules; moderator condition incoherent; `decision_out`/`signal_out` payloads missing); Claude **D3** (moderator failure path).
- **Problem:** `ModuleAssistanceRequest` has no endpoint/lease/timeout/answer schema; participant policy can't model non-module participants; the moderator-required condition is incoherent; `decision_out`/`signal_out` payloads are undefined; and there's no behavior when the moderator agent is unavailable (D3).
- **Disposition:** **ACCEPT.** Adopt the full `ModuleAssistanceRequest` (Appx H) + a moderator fallback analogous to the Task Agent fallback.
- **Fix -- `[Appx H]`:**
```ts
// ModuleAssistanceRequest { target, target_endpoint_ref, answer_schema_ref, request_kind, response_policy,
// lease {holder_ref, lease_version, expires_at}, timeout_ms, on_timeout: resume_with_warning|abort|escalate_human|ask_task_agent }.
// Participant policy models module|subagent|task_agent|user. decision_out/signal_out get explicit payload schemas.
// Sec. 4.6A moderator fallback: moderator unavailable -> degrade to "none" | pause | queue (default: ask_task_agent then pause).
```
### D-11 -- [Run Board Sec. 3.1 / Sec. 6.2] BoardDigest filter rule unspecified (ACCEPT) -- Claude C9
- **Raised by:** Claude **C9**.
- **Problem:** `BoardDigest` carries `included_post_ids` but the **selection rule** is unspecified -- a 500-post forum can't ship all posts in a ~1200-token digest, and implementations pick differently.
- **Disposition:** **ACCEPT.** Extend `BoardDigestPolicy` with explicit selection.
- **Fix:**
```ts
// BoardDigestPolicy adds: included_post_kinds: TaskRunBoardPostKind[]; included_severity_threshold;
// max_posts: number; selection_strategy: "recency"|"severity"|"score"|"mixed".
// Default: kinds {evaluation_finding, repair_instruction, process_gap, user_guidance}; severity >= medium; max 30; mixed.
```
### D-12 -- [Run Board Sec. retention / Source Workspace Sec. 9] Run Board retention/compaction/event-class + EC-policy persistence (ACCEPT)
- **Raised by:** ChatGPT (retention/compaction/event-class missing; persistence should reference EC policy).
- **Problem:** the Run Board has no retention, compaction, or event-class taxonomy, and persistence doesn't reference the EC policy decisions that should govern it.
- **Disposition:** **ACCEPT.** Add retention + compaction policies and an event-class taxonomy; bind persistence to EC policy.
- **Fix:**
```ts
type RunBoardEventClass = "post" | "digest" | "assistance_request" | "moderation" | "lifecycle";
interface RunBoardRetentionPolicy { retain_event_classes: RunBoardEventClass[]; retain_for_days: number;
ec_policy_decision_ref: PolicyEvaluationRef; }
interface RunBoardCompactionPolicy { compact_after_days: number; strategy: "summarize"|"drop_low_severity"|"archive"; }
```
**D-13** -> rendered in Layer 3 core (that rendering governs).
### D-14 -- [Core Sec. 3D / Sec. 4B / Sec. 5A] InjectionSlotRegistry, compact card schemas, command registry, token_budget, DOC24 capability ownership, receipt booleans (ACCEPT) -- incl. Claude C15
- **Raised by:** ChatGPT (InjectionSlotRegistry gap CRITICAL; compact card schemas missing; Core defines a DOC24-owned capability; `TrackedTaskReceipt` booleans need command refs; command registry should be mandatory); Claude **C15** (packet taxonomy unstated); Grok GK-5.11 (`token_budget` never populated).
- **Problem:** the `InjectionSlotRegistry` for the task-system DOC24 slots is unspecified; the compact top-k card schemas are missing; Core normatively defines a DOC24-owned capability; `TrackedTaskReceipt` carries action booleans without command refs; `TaskOpportunityPacket.token_budget` is never populated/checked; and the packet taxonomy (opportunity vs run-context vs design) is unstated.
- **Disposition:** **ACCEPT.** Adopt the Core appendix wholesale; move the capability definition to DOC24; make the command registry mandatory.
- **Fix -- `[Appx I]`:**
```ts
// TaskSystemInjectionSlotRegistration { slot_id (6 DOC24 slots), slot_kind, surfaces, token_cap,
// pii_redaction_required, on_unavailable: omit|degrade_direct_first|block_explicit_task_route, receipt_required }.
// Compact cards: CompactTaskInvocationDirectiveCard / CompactTaskTemplateCard / CompactModulePresetCard
// (each with CalibratedScore, risk_flags, token_estimate, redaction_state, source_authority).
// TaskCommandRegistryEntry (mandatory) { request/response_schema_ref, idempotency_key_required, durable_write,
// telemetry_event_kind, read_model_invalidations, failure_codes, required_policy_checks, owning_doc }.
// TrackedTaskReceipt.available_actions: AvailableTaskAction[] (each -> command_ref + idempotency_key_required).
// Packet taxonomy subsection (C15): TaskOpportunityPacket (pre-task), TaskRunContextPacket (in-run), TaskAgentDesignPacket (future).
// Capability def moves to DOC24 (Core references). token_budget populated + checked at assembly.
```
### D-15 -- [V3.3.1 Sec. 8.4 / Sec. 17.1] Sub-agent: output-contract plurality, no-sub-agent fallback, coordination-point count (ACCEPT) -- incl. Claude B10, D1
- **Raised by:** ChatGPT (`output_contract_ref` singular vs variant semantics; no-sub-agent fallback incomplete); Claude **B10** (no fallback at evaluator point) + **D1** (coordination point count 4 vs 5).
- **Problem:** `output_contract_ref` is singular but variant semantics need plurality; the "no sub-agent available" fallback is incomplete (esp. at the evaluator coordination point); and Sec. 17.1 lists five coordination points while Sec. 8.4's enum has four (no `plan_verifier`), so a profile can't declare for the fifth.
- **Disposition:** **ACCEPT.** Plural output contracts; a fallback policy per coordination point; reconcile to five points.
- **Fix:**
```ts
// Sec. 8.4 allowed_coordination_points += "plan_verifier" (reconcile to Sec. 17.1's five).
// output_contract_refs: StorageRef[] (was singular).
interface SubAgentFallbackPolicy { coordination_point: AllowedCoordinationPoint;
on_no_sub_agent: "use_primary_module" | "skip_with_warning" | "hard_call" | "degrade_quality_with_note"; }
// every coordination point declares a fallback; validation.sub_agent_point_without_fallback (error).
```
### D-16 -- [Source Workspace / Forum] Workspace API operations referenced but never defined (ACCEPT) (!)verify
- **Raised by:** Grok GK-4.2.
- **Problem:** workspace operations (create/read/append/lock) are referenced but never defined -- a circular reference with no API surface.
- **Disposition:** **ACCEPT.** Define the workspace API (operations, args, receipts), bound to the command registry (D-14) and write-precondition (B-03).
- **(!)verify:** confirm the current state of any partial workspace-API definition before authoring.
### D-17 -- [Common Contracts Sec. 7 / V3.3.1 Sec. 7.9.3] Anchor/hash hygiene: empty StructuredAnchor, uncomputed context_hash, HardCall hash normalization (ACCEPT) -- incl. Claude C7, D14
- **Raised by:** Grok GK-5.5 (`TextAnchor.context_hash` never computed/validated); Claude **C7** (`StructuredAnchor {}` validates but is useless) + **D14** (`HardCallResolution` hash normalization unspecified).
- **Problem:** `context_hash` is defined but never computed/validated; a `StructuredAnchor` with all-optional fields can be validly empty yet un-resolvable; and `HardCallResolution` reuse compares hashes whose normalization is unspecified (cosmetic diffs -> needless re-escalation).
- **Disposition:** **ACCEPT.** Compute/validate anchors; require non-empty structured anchors; specify canonical normalization before hashing.
- **Fix:**
```
StructuredAnchor MUST populate >=1 of {section_id, field_path, citation_ref}; validation.structured_anchor_empty (error).
TextAnchor.context_hash MUST be computed at creation and validated at resolve; validation.context_hash_unverified (warning).
HardCallResolution: canonicalize outcome_definition_hash / goal_context_hash inputs (trim, sort keys, normalize whitespace)
before hashing (Sec. 7.9.3); validation.hard_call_hash_unnormalized.
```
### D-18 -- [V3.3.1 Sec. repeated-failure] Repeated-failure detection keyed on versioned refs (ACCEPT)
- **Raised by:** Grok GK-5.12.
- **Problem:** repeated-failure detection keys on `affected_artifact_refs`, which are version-stamped, so the same logical artifact across versions doesn't match -- repeats go undetected.
- **Disposition:** **ACCEPT.** Key on stable artifact identity, not the versioned ref.
- **Fix:** detection key = stable `artifact_id` (+ `outcome_definition_id`), not `artifact_version_ref`; `validation.repeated_failure_keyed_on_version` (lint).
### D-19 -- [Source Workspace Sec. SourceRecord] Per-module cost attribution has no field (ACCEPT)
- **Raised by:** Grok GK-5.10.
- **Problem:** cost can't be attributed per module on `SourceRecord` -- no field -- so per-module cost reporting (C-07 / BudgetNarrative) has nothing to read.
- **Disposition:** **ACCEPT.** Add a per-module cost field.
- **Fix:** `SourceRecord.acquisition_cost?: CostEstimate` (C-07 type) + `acquired_by_module_id`; feeds the cost rollups.
### D-20 -- [Source Workspace Sec. library] Library promotion gate references EC policy but is undefined (ACCEPT)
- **Raised by:** Grok GK-4.10.
- **Problem:** promotion of a workspace source into the durable Library references an EC policy gate, but the gate itself isn't defined.
- **Disposition:** **ACCEPT.** Define the promotion gate (criteria + policy check + receipt).
- **Fix:** `LibraryPromotionGate { min_tier, requires_verification_state, ec_policy_decision_ref, requires_access_tier }`; promotion emits a receipt; `validation.library_promotion_without_gate` (error).
### D-21 -- [V3.3.1 / Core] `requires_background_progress` overloaded (ACCEPT)
- **Raised by:** ChatGPT (CG audit addendum).
- **Problem:** `requires_background_progress` is used for two different meanings (a run needing background work vs a module needing periodic progress) with no disambiguation.
- **Disposition:** **ACCEPT.** Split into two named fields.
- **Fix:** replace with `requires_background_execution: boolean` (run-level) and `emits_progress_heartbeat: boolean` (module-level); migrate references.
### D-22 -- [Common Contracts Sec. 11.5 / cross-refs] Backward-compat overstated; section-anchor hygiene before R3.2 absorption (ACCEPT WITH MODIFICATIONS) -- Claude D5/D22
- **Raised by:** Claude **D5** + **D22**.
- **Problem:** Sec. 11.5's backward-compat claim overstates stability, and the set uses section-number cross-refs where stable anchors are needed before the DOC23 R3.2 absorption. (Overlaps A-15; tracked there for the build-ready gate.)
- **Disposition:** **ACCEPT WITH MODIFICATIONS.** Soften the compat claim to "compatible within the locked schema-of-record set; no cross-version guarantee yet"; convert section-number cross-refs to stable anchors as pre-absorption hygiene.
- **Fix:** anchor-stabilization pass across the set before R3.2 absorption; `validation.unstable_section_cross_ref` (lint). Coordinate with A-15.
### D-23 -- [FD Sec. 9.4] "Silent ignoring fires validation" is unenforceable (ACCEPT) -- Claude C11
- **Raised by:** Claude **C11**.
- **Problem:** Sec. 9.4 says silent ignoring (no receipt) fires `validation.feedback_consumed_without_receipt` at audit, but the audit has no way to enumerate *expected* receipts -- so it never fires; a module that processed-and-ignored looks identical to one that received nothing.
- **Disposition:** **ACCEPT.** Track receipt expectations at dispatch so the audit can compare.
- **Fix:**
```ts
// Sec. 9.4A: when the router (Sec. 6) dispatches a bundle, record a FeedbackDispatchExpectation keyed to
// (feedback_bundle_id, consumer_module_id, consumer_activation_seq). At run end (or after a grace period,
// default 5 min post-activation) the audit compares expectations <-> FeedbackConsumptionReceipts [Appx F];
// missing pairs fire validation.feedback_consumed_without_receipt.
```
**D-24** -> rendered in Layer 3 core (that rendering governs).
### R.5 -- Sec.E rows (held, Phase-B-gated)
# Sec. E. HELD -- Phase-B-gated / self-learning (tracked, not adjudicated for build)
Per your instruction: held but tracked. Gated on the Phase-B corpus audit (writing these now = guessing their own spec).
- **E-01 Memory Hydration phase** -- pre-run `HydrateMemory` step (query BDSM/DOC72/RunGuidance, resolve by precedence, inject priors, stamp `HydratedMemoryHash`). *Architectural absence; adopted in principle, build gated.* (Gemini Sec. 7; Claude Sec. 6.1/Sec. 6.3)
- **E-02 Memory Precedence Hierarchy** -- `Local Intent > Matter/Scope Policy > Global DOC72`. (Gemini Sec. 7; Claude Sec. 6.1/Sec. 6.3)
- **E-03 SubAgentPrior injection** ("Sub-Agent Amnesia") -- inject last-N relevant findings/RepairInstructions per sub-agent. (Gemini GM-V1; Claude Sec. 6.1/Sec. 6.3)
- **E-04 TaskBlueprint Topology/Payload bifurcation** -- `TaskTopology` (generalized) vs `TaskPayload` (ephemeral). (Gemini Sec. 7; Claude Sec. 6.1/Sec. 6.3)
- **E-05 Flawless-execution denominator** (learning side) -- count clean passes via verdict-aware `OutcomeEvaluationSignal` so utility doesn't falsely decay. *(This is the real home of the deleted `EvaluationAffirmation`, A-09.)* (Gemini GM-V2; Claude Sec. 6.1)
- **E-06 Sub-agent reputation** model/slices/model-class calibration (C-04's weighting feeds this). (Claude D22; ChatGPT CG-SA3/CG-SL12)
- **E-07 `LearningMode` consumption behavior** (A-11's value set is adopted; the behavior is held); `goal_advancement_count` decrement (D4); `cross_model_applicability` runtime behavior (D21).
- **E-08 Process-gap -> design-pattern loop** (A6) + substantive-vs-process gap enforcement. (Claude A6; Grok GK-3.5)
- **E-09 Longitudinal pattern view (S6)** and other learning-touching surfaces.
- **E-10 All ChatGPT `[SELF-LEARNING]` items (CG-SL1..12)** incl. TIE / Task Improvement Engineer (**Appendix P**, held), LoopEffectivenessTestRunRecord, BDSM utility-compilation gap, revealed-preference dampening, outcome clustering, UserConstitution prior, multi-prior conflict, PlanDiff/ProposalDiff, active-learning bundle.
- **E-11 Design-Feedback as peer to Artifact-Repair** -- elevate Task Agent graph-patch proposals to first-class learned outcomes. (Grok GK-3.2/N1)
- **E-12 Automatic Pattern Suggestion UI** (>=2 failures -> top-3 similar patterns). (Grok GK-N4)
---
### E-13 -- [Common Contracts learning envelope] Multi-user forward-compatibility fields (DEFER-PhaseB; fields addable now)
- **Raised by:** ChatGPT (GAP/HIGH); Claude self-learning analysis.
- **Problem:** the learning-signal envelope has `data_class` and `matter_id` but lacks `principal_id`, learning scope, share eligibility, and scope-inference basis. Retrofitting team/firm/networked learning scope later -- over privilege and matter data -- is dangerous.
- **Disposition:** **DEFER-PhaseB for behavior; fields addable now as forward-compat** (same pattern as A-11 value-set / C-04 math: adopt the schema surface now, defer the multi-user learning *behavior* to Phase B). Flagged here so it isn't lost; the actual networked-learning semantics are part of the Phase-B learning spec.
- **Fix (forward-compat fields, addable now):**
```ts
// add to learning signal envelope / learning artifacts:
principal_id: string;
learning_scope: "local" | "matter" | "team" | "firm" | "networked";
scope_inference_basis: string;
default_scope_rule: string;
share_eligibility: "none" | "opt_in" | "policy_gated";
```
---
---
# Sec. F. DECLINED -- review Sec. 6.2 (closed; not re-opened)
| # | Proposed | Raised by | Why declined / adopted-instead |
|---|---|---|---|
| F-01 | Literal Git-style branching / ShadowWorkspace as core primitive (Branch & Merge) | Gemini headline, Grok GK-N2 | Side effects can't branch (phantom). **Instead:** `TaskRunFork` + `irrevocable_side_effects_at_fork`. |
| F-02 | `TaskConfirmationSignal` as a new signal type | Gemini Sec. 7 | Duplicates existing signal/receipt machinery. **Instead:** fold into existing signals/receipts. |
| F-03 | Flawless-execution as a new signal type | Gemini GM-V2 | Redundant. **Instead:** verdict-aware `OutcomeEvaluationSignal` (denominator need kept -- see E-05). |
| F-04 | Chunking findings for KV-cache bloat | Gemini D-02 | Fragments the unit. **Instead:** compressed-envelope view at prompt-assembly. |
---
---
# Sec. G. Conceptual / UX / surfaces / Professional Reliance Layer
> **Why this cluster matters (plain language):** Sec. A-Sec. D make the engine *correct*. Sec. G is what makes the output something a litigator can actually *rely on and hand to a partner* -- a pre-flight check on what the system thinks you asked, a reviewable diff of every change, an evidence binder tying each claim to its source, a plain-English cost/quality account, and a single "can I rely on this, and within what limits" cover memo. The review calls this the biggest product opportunity in the set, and it's the part that differentiates ELNOR from a generic agent in 2027.
### R.6 -- Sec.F rows (declined)
# Sec. F. DECLINED -- review Sec. 6.2 (closed; not re-opened)
| # | Proposed | Raised by | Why declined / adopted-instead |
|---|---|---|---|
| F-01 | Literal Git-style branching / ShadowWorkspace as core primitive (Branch & Merge) | Gemini headline, Grok GK-N2 | Side effects can't branch (phantom). **Instead:** `TaskRunFork` + `irrevocable_side_effects_at_fork`. |
| F-02 | `TaskConfirmationSignal` as a new signal type | Gemini Sec. 7 | Duplicates existing signal/receipt machinery. **Instead:** fold into existing signals/receipts. |
| F-03 | Flawless-execution as a new signal type | Gemini GM-V2 | Redundant. **Instead:** verdict-aware `OutcomeEvaluationSignal` (denominator need kept -- see E-05). |
| F-04 | Chunking findings for KV-cache bloat | Gemini D-02 | Fragments the unit. **Instead:** compressed-envelope view at prompt-assembly. |
---
---
### R.7 -- Sec.G rows (full schemas; the Layer 4 classification governs each row's disposition/landing)
# Sec. G. Conceptual / UX / surfaces / Professional Reliance Layer
> **Why this cluster matters (plain language):** Sec. A-Sec. D make the engine *correct*. Sec. G is what makes the output something a litigator can actually *rely on and hand to a partner* -- a pre-flight check on what the system thinks you asked, a reviewable diff of every change, an evidence binder tying each claim to its source, a plain-English cost/quality account, and a single "can I rely on this, and within what limits" cover memo. The review calls this the biggest product opportunity in the set, and it's the part that differentiates ELNOR from a generic agent in 2027.
## Professional Reliance Layer (Appendix O + M)
### G-01 -- [new - Common Contracts / Core] `EvaluationContractReview` -- pre-execution contract check (ACCEPT)
- **Raised by:** ChatGPT (IDEA/CRITICAL).
- **What & why:** before the system spends tokens/time, it surfaces *what it thinks you asked for and how it will judge success* -- interpreted goal, criteria, thresholds, source requirements, required capabilities, and the Hard-Call triggers -- and lets you approve, edit, or waive. Catches a misread brief before the cost is incurred (the cheapest possible place to catch it). **User-facing:** a short "here's the plan and the bar; approve to proceed" card.
- **Disposition:** **ACCEPT** as a new pre-execution artifact (gate optional per task autonomy).
- **Fix -- `[Appx O]`:**
```ts
interface EvaluationContractReview {
review_id: string; task_id: string; run_id?: string; compiled_plan_ref: StorageRef;
interpreted_goal: string; criteria_summary: string[]; threshold_summary: string[];
source_requirements: string[]; required_capabilities: CapabilityRef[]; hard_call_triggers: HardRevisionCallKind[];
material_differences_from_preview?: string[];
user_approval_required: boolean;
approval_status: "pending" | "approved" | "rejected" | "edited" | "waived_by_policy"; approval_ref?: StorageRef;
created_at: ISO8601; schema_version: "1.0";
}
```
### G-02 -- [new - FD / V3.3.1] `RevisionReviewPacket` -- reviewable packet for every meaning-bearing revision (ACCEPT)
- **Raised by:** ChatGPT (IDEA/HIGH).
- **What & why:** every meaning-bearing revision emits a packet showing the before->candidate semantic diff, which finding drove which change (`finding_to_change_map`), preservation-constraint results, source changes, and revalidation results -- so a reviewer can accept/reject/fork/request-changes/restore *with the reasoning visible*, not just a new blob. **User-facing:** "here's exactly what changed and why; accept, fork, or roll back."
- **Disposition:** **ACCEPT.** Pairs with G-04 (restore) and A-16 (revision goes through `revision_in`).
- **Fix -- `[Appx O]`:**
```ts
interface RevisionReviewPacket {
packet_id: string; task_id: string; run_id: string; revision_plan_ref: StorageRef;
before_artifact_version_ref: StorageRef; candidate_artifact_version_ref: StorageRef;
semantic_diff_ref: StorageRef; finding_to_change_map: Record<string, string[]>;
preservation_constraint_result_refs: StorageRef[]; source_changes: SourceRecordRef[]; revalidation_result_refs: StorageRef[];
regression_risk_summary: string;
reviewer_action: "accept" | "reject" | "fork" | "request_changes" | "restore_known_good_state" | "no_user_review_required";
created_at: ISO8601; schema_version: "1.0";
}
```
**Audit fold-in (R0.4) -- review decision receipt** (DEEP Sec.8 / ChatGPT Sec.11). *Resolves the prior inline-`reviewer_action` divergence:* the packet keeps `reviewer_action` for convenience, but the durable record of a decision is a receipt, so a meaning-bearing revision cannot ship unreviewed.
```ts
interface RevisionReviewDecisionReceipt {
receipt_id: string; packet_id: string; // -> RevisionReviewPacket
decision: "approved" | "rejected" | "approved_with_conditions";
reviewer_ref: string; conditions?: string[]; decided_at: ISO8601;
governance: GovernanceEnvelope; schema_version: 1;
}
// lints: validation.review_decision_without_receipt ; validation.revision_review_packet_contains_current_action ; validation.meaning_bearing_revision_without_review_packet_or_policy_waiver
```
### G-03 -- [new - Source Workspace] `EvidencePackage` -- exportable evidence binder (ACCEPT)
- **Raised by:** ChatGPT (IDEA/HIGH).
- **What & why:** the Source Workspace exported as one reviewable unit -- final artifacts + workspace snapshot + a `claim_support_map` marking each claim `supported / partially_supported / unsupported / contradicted / not_checked`, plus unresolved research needs and stale/unverified sources. This is the binder a litigator hands to a partner or keeps for the file. **User-facing:** "every claim, what backs it, and what's still unverified." Depends on D-06 evidence anchors.
- **Disposition:** **ACCEPT.**
- **Fix -- `[Appx O]`:**
```ts
interface EvidencePackage {
evidence_package_id: string; task_id: string; run_id: string;
final_artifact_refs: StorageRef[]; source_workspace_snapshot_ref: StorageRef;
source_record_refs: SourceRecordRef[]; evidence_anchor_refs: string[];
claim_support_map: Array<{ claim_ref: ClaimRef; supporting_anchor_refs: string[]; contradicting_anchor_refs: string[];
support_status: "supported" | "partially_supported" | "unsupported" | "contradicted" | "not_checked"; }>;
unresolved_research_need_refs: string[]; stale_or_unverified_source_refs: SourceRecordRef[];
created_at: ISO8601; schema_version: "1.0";
}
```
### G-04 -- [new - V3.3.1 / Core] `KnownGoodState` -- named restorable checkpoint (ACCEPT)
- **Raised by:** ChatGPT (IDEA/HIGH).
- **What & why:** a named, restorable checkpoint of a run/artifact state, so cancel (B-10), fork (G-17), and revision-review (G-02) all have a concrete thing to roll back to. **User-facing:** "save point -- restore if a later change makes things worse."
- **Disposition:** **ACCEPT.** Schema in Appendix M; `restore_known_good_state` is already an `AvailableTaskAction` (D-14) and a `reviewer_action` (G-02).
- **Fix -- `[Appx M]`:** `KnownGoodState { state_id, task_id, run_id, artifact_version_refs[], workspace_snapshot_ref, label, created_at }` + a `restore` command in the command registry (D-14) with an idempotency key.
**Audit fold-in (R0.4) -- `KnownGoodCheckpoint` hardening** (DEEP Sec.8 / ChatGPT Sec.11). The base `KnownGoodState` is hardened into a checkpoint that snapshots policy + capability context, so a restore cannot silently re-expose or re-enable.
```ts
interface KnownGoodCheckpoint {
checkpoint_id: string;
artifact_version_refs: StorageRef[];
policy_snapshot: PolicyEvaluationRef; // policy generation in force at checkpoint
capability_snapshot: CapabilityRef[]; // capabilities granted at checkpoint
irreversible_side_effects_before: string[]; // effects a restore CANNOT undo
governance: GovernanceEnvelope; // L1.1
schema_version: 1;
}
// lints: validation.known_good_checkpoint_missing_policy_snapshot ; validation.known_good_checkpoint_missing_capability_snapshot ; validation.restore_checkpoint_ignores_irreversible_side_effects
```
### G-05 -- [new - Core / Common Contracts] `BudgetNarrative` -- plain-English cost/quality account (ACCEPT)
- **Raised by:** ChatGPT (IDEA/HIGH).
- **What & why:** a human-readable account of what a run cost and what that bought -- separating *logical* LLM calls from *infrastructure retries* (so retries don't look like work), local compute, external tool cost, which optional helpers were skipped, which non-degradable modes were preserved, which degraded modes were used, and the quality impact. **User-facing:** "what it cost, what was skipped, and how that affected quality."
- **Disposition:** **ACCEPT.** Consumes the shared `CostEstimate`/`TaskCostRecord` (C-07) and feeds A-02 / G-09 cost predictability.
- **Fix -- `[Appx O]`:**
```ts
interface BudgetNarrative {
budget_narrative_id: string; task_id: string; run_id: string;
planned_estimate_ref?: StorageRef; actual_cost_records: TaskCostRecord[]; summary: string;
logical_llm_calls: number; infrastructure_retries: number; local_compute_seconds: number;
external_tool_cost_usd?: number; doc24_packet_assembly_cost?: CostEstimate; source_research_cost?: CostEstimate;
skipped_optional_helpers: string[]; preserved_non_degradable_modes: string[]; degraded_modes_used: string[];
quality_impact_summary: string; created_at: ISO8601; schema_version: "1.0";
}
```
### G-06 -- [new - Common Contracts / Core] `TaskReliancePacket` -- the capstone reliance artifact (ACCEPT)
- **Raised by:** ChatGPT (IDEA/CRITICAL).
- **What & why:** the cover memo that ties the whole layer together -- assurance summary, unresolved limitations, the evidence package, revision-review packets, Hard-Call resolutions, policy decisions, budget narrative, and known-good states -- and renders a single **`reliance_status`**: `safe_to_rely_within_scope` / `rely_with_limitations` / `not_safe_to_rely` / `human_review_required`, with an explicit `reliance_scope` and a user-visible summary. This is the artifact a high-stakes professional actually relies on. **User-facing:** "can I rely on this, within what scope, with what caveats."
- **Disposition:** **ACCEPT** -- the keystone of Sec. G; everything else in this cluster feeds it.
- **Fix -- `[Appx O]`:**
```ts
interface TaskReliancePacket {
packet_id: string; task_id: string; run_id: string;
final_artifact_refs: StorageRef[]; evaluation_chain_ids: string[]; assurance_summary_ref: StorageRef;
unresolved_limitations: EvaluationLimitationKind[]; // <- A-07 limitation taxonomy
evidence_package_ref?: StorageRef; revision_review_packet_refs: StorageRef[]; hard_call_resolution_refs: StorageRef[];
policy_decision_refs: PolicyEvaluationRef[]; budget_narrative_ref?: StorageRef; known_good_state_refs: string[];
reliance_status: "safe_to_rely_within_scope" | "rely_with_limitations" | "not_safe_to_rely" | "human_review_required";
reliance_scope: string; user_visible_summary: string; created_at: ISO8601; schema_version: "1.0";
}
```
- **Rationale:** binds the limitation taxonomy (A-07), evidence anchors (D-06), Pattern C chains (A-03), and budget (C-07) into one auditable statement. Without it, "the system did good work" is unverifiable; with it, reliance is scoped and inspectable.
### G-07 -- [new - DOC20 / Core] `AttentionLedger` / `DecisionQueue` -- cross-run attention surface (ACCEPT -- author minimal)
- **Raised by:** ChatGPT (IDEA/HIGH).
- **What & why:** one place that surfaces everything across tasks/runs awaiting the user's attention or decision -- pending Hard Calls, blocked items, contested findings, approvals. Prevents decisions from being buried inside individual runs. **User-facing:** "one inbox of everything that needs your call."
- **Disposition:** **ACCEPT.** No Appendix schema; author a minimal one for review.
- **Fix (PROPOSED -- confirm shape):**
```ts
interface AttentionLedgerItem {
item_id: string; task_id: string; run_id?: string;
attention_kind: "hard_call_pending" | "blocked_item" | "contested_finding" | "approval_required" | "research_need_human";
summary: string; priority: "low" | "medium" | "high" | "blocking";
command_ref?: string; created_at: ISO8601; resolved_at?: ISO8601; schema_version: "1.0";
}
```
### G-08 -- [Set-wide] No first-class "task health" surface (ACCEPT) -- Claude A1
- **Problem/why:** task health is fragmented across signals; there's no single surface showing progress, blockers, budget burn, and last verdict. **Disposition:** **ACCEPT** -- define a `TaskHealthCard` DOC20 surface aggregating the existing signals (no new truth). Renders from run state + BudgetNarrative (G-05) + last `OutcomeEvaluationState`.
### G-09 -- [Set-wide] Cost predictability asserted but not computable before a run (ACCEPT) -- Claude A2
- **Problem/why:** the set claims cost predictability but nothing assembles an end-to-end forecast before running. **Disposition:** **ACCEPT** -- a pre-run forecast built from the shared `CostEstimate` (C-07) summed across packets/research/LLM calls (the D-18 per-activation budget gives the sub-totals); surfaced in EvaluationContractReview (G-01) and reconciled after by BudgetNarrative (G-05).
### G-10 -- [V3.3.1 Sec. 21 / FD Sec. 8] Reviewability fragmented across >=3 surfaces (ACCEPT) -- Claude A3
- **Problem/why:** review is scattered across at least three places. **Disposition:** **ACCEPT** -- unify into one review experience: FindingsInbox (G-13) for triage, RevisionReviewPacket (G-02) for changes, DecisionAuditView (G-15) for the "why." One entry point, not three.
### G-11 -- [FD Sec. 3.4 / V3.3.1 Sec. 6.12] Over-relies on the user knowing to contest (ACCEPT) -- Claude A5
- **Problem/why:** findings are defeasible (FD Sec. 3.4) but the UI doesn't *surface* contestability, so the burden is on the user to know they can push back. **Disposition:** **ACCEPT** -- proactively mark contestable findings/verdicts in the UI with the contest affordance inline (ties to the defeasible-findings model).
### G-12 -- `WorkProductCertification` -- the page you staple to the cover sheet (ACCEPT -- highest-leverage surface) -- Claude S1
- **What:** the human-facing render of `TaskReliancePacket` (G-06): reliance status, scope, limitations, evidence summary. **Disposition:** **ACCEPT.** DOC20 surface over G-06; read-only; no new truth.
### G-13 -- `FindingsInbox` -- cross-task review queue (ACCEPT) -- Claude S2
- **What:** a queue of findings/repair instructions across tasks, filterable by severity/state/matter. **Disposition:** **ACCEPT.** Reads canonical `EvaluationFinding`s (A-01); matter-scoped per D-24.
### G-14 -- `RunDiff` -- compare two runs of the same task (ACCEPT) -- Claude S3
- **What:** diff two runs (inputs, plan, outcomes, cost). **Disposition:** **ACCEPT.** Reads run records + BudgetNarrative (G-05); pairs with RunReplayPreview (G-16).
### G-15 -- `DecisionAuditView` -- "why did it decide that" (ACCEPT) -- Claude S4
- **What:** renders the decision chain behind a verdict/route (which findings, which policy, which Hard Call). **Disposition:** **ACCEPT.** The UI counterpart to the Pattern C chain (A-03) and routing (A-08); renders the coordination trace.
### G-16 -- `RunReplayPreview` -- preview a replay before committing (ACCEPT) -- Claude S5
- **What:** show what a replay would produce before applying it. **Disposition:** **ACCEPT.** Depends on the TaskReplay primitive (G-19) + KnownGoodState (G-04).
### G-17 -- [Core R0.7.1 Sec. 5.1] `TaskRunFork` + `irrevocable_side_effects_at_fork` (ACCEPT -- the adopted form of the declined ShadowWorkspace) -- Claude Sec. 5.1
- **What & why:** fork a run from a checkpoint to explore an alternative without disturbing the original -- the *real* answer to Gemini/Grok's branch-and-merge idea (Sec. F-01), made honest by an explicit record of side effects that **cannot** be forked (an already-sent email stays sent). **Disposition:** **ACCEPT.**
- **Fix (PROPOSED, consistent with Sec. 5.1):**
```ts
interface TaskRunFork {
fork_id: string; parent_run_id: string; forked_from_checkpoint_ref: string; // KnownGoodState (G-04)
new_run_id: string; fork_reason: string;
irrevocable_side_effects_at_fork: Array<{ side_effect_id: string; kind: string; executed_at: ISO8601;
note: "not_reversible_in_fork"; }>;
created_at: ISO8601; schema_version: "1.0";
}
```
### G-18 -- [V3.3.1 / Core] `ExplanationTrace` as a first-class artifact (ACCEPT) -- Grok
- **What & why:** every `CompiledRevisionStrategy`/`RevisionPlan` emits a short human-readable causal trace ("changed X because finding Y, preserving Z"). Cheap to produce, large trust gain, and feeds DecisionAuditView (G-15). **Disposition:** **ACCEPT** -- a short `explanation_trace_ref` on the plan/strategy.
### G-19 -- [Set-wide] `TaskReplay` primitive (ACCEPT) -- Claude D11
- **What & why:** deterministic replay of a run closes the determinism story and underpins RunReplayPreview (G-16) and RunDiff (G-14). **Disposition:** **ACCEPT** -- a replay primitive keyed to a run snapshot + KnownGoodState (G-04).
**Audit fold-in (R0.4) -- replay request + divergence** (DEEP Sec.8 / Claude D11). The G-19 concept above gets its schemas; a replay must never re-fire irreversible external side effects.
```ts
interface TaskReplayRequest {
replay_id: string; source_run_ref: string; mode: "deterministic" | "best_effort";
side_effect_policy: "suppress_external" | "sandbox_only"; schema_version: 1;
}
interface ReplayDivergenceRecord {
replay_id: string; step_ref: string;
divergence_kind: "nondeterministic_step" | "input_changed" | "policy_changed" | "tool_unavailable"; detail: string;
}
// lints: validation.replay_attempts_external_side_effect ; validation.deterministic_replay_with_nondeterministic_step ; validation.replay_divergence_unrecorded
```
### G-20 -- [DOC20] Unified Evaluation-Chain view (ACCEPT) -- Grok
- **What & why:** render a Pattern C chain as one card -- qualitative findings and the quantitative verdict side by side -- instead of two disconnected envelopes. The UI counterpart to A-02/A-03/A-04. **Disposition:** **ACCEPT** -- DOC20 surface over the chain registry (A-03).
### G-21 -- [Testing] Chaos / concurrency fixtures (ACCEPT) -- ChatGPT / Claude Sec. 6.3
- **What & why:** the test harness for the Sec. B concurrency work -- storage-full, malformed LLM output, mid-run privilege change, clock skew, parallel writes to one artifact. **Disposition:** **ACCEPT** -- required fixtures backing B-03/B-04/B-05/B-16 and D-01/D-24; gate the runtime fixes on them passing.
---
---
# Sec. G/Sec. A/Sec. D partials -- folded as notes (not full rows)
Covered in substance by existing rows; recorded here so they're tracked, with where each attaches:
- **`authority_basis` array can be empty while hard blockers require backed authority** (Grok Sec. 5.2) -> note on **A-05**: add a non-empty constraint -- a finding with `severity:"blocking"` MUST carry >=1 `assurance_basis`; `validation.blocking_finding_without_authority` (error).
- **`AutonomousModePolicy` as single source of truth + visible toggle/live risk score** (Grok Sec. 3.1) -> note on **A-16 / G-08**: Claude D9 already confirms the *locked fields* are correct-by-construction; add the UX requirement to surface `AutonomousModePolicy` as a visible toggle with a live risk score, and the invariant that no mutation path bypasses it (reinforces A-16).
- **Consumption-receipt -> `RepairCycleSignal` linkage is optional/never mandated** (Grok Sec. 4.6) -> note on **D-23**: make the `consumption_receipt_ref -> RepairCycleSignal` linkage field **required**, not optional; `validation.repair_cycle_signal_without_consumption_link` (error).
- **OP-A rows claim the same primitive under different names** (Grok Sec. 4.4) -> note on **A-15 / supersession (Sec. 8)**: the supersession matrix + TypeOwnerRegistry already force one canonical name per primitive; add an OP-A dedup pass as a build-ready gate item.
- **Mandatory Plan-Verifier for high-risk plans** (Grok top-5) -> note on **D-15**: D-15 adds the `plan_verifier` coordination point; add the *policy* that a `plan_verifier` sub-agent is **required** (not optional) when `plan_risk_score >= threshold`.
---
---
*Consolidated from working cards V0.1-V0.6. 126 items adjudicated from ~228 raw reviewer assertions: 88 fixes (Sec. A-Sec. D), 13 held for Phase B (Sec. E), 4 declined per review Sec. 6.2 (Sec. F), 21 product surfaces (Sec. G), plus 5 partials folded as notes. Fixes carry the Appendix A-P schemas inline. `(!)verify` items deferred for current-state confirmation. Next step after your decisions: the R0.4 amendment package applying the accepted rows to the operative specs.*
---
## Final consistency pass
**Coverage (126 + new):** every consolidated-card item lands in exactly one place -- Sec. A 27 (L3 core 8 - L2 2 - standard 17), Sec. B 27 (L3 core 5 - L2 2 - standard 20), Sec. C 10 (L3 core 1 [C-03] - C-cluster 9), Sec. D 24 (L3 core 2 - L2 1 - standard 21), Sec. G 21 (Layer 4; G-21->Layer 5), Sec. E 13, Sec. F 4 = **126**. New rows NR.1-NR.7 add scope (NR.4 completes the D-19/D-20 standard rows). No item dropped; flips and re-grades explicit.
**Primitives owned:** all seven L1 primitives have `TypeOwnerRegistry` entries (NR.1); no later row redefines them.
**Lints -- all registered in `LintRegistry` (L1.7):** Step-0 (x5), Layer-2 (x3), Layer-3 core (x6), C-cluster (x4), Layer-4 (x4), Layer-5 fixtures (their asserted lints), new-rows (reliance/delegation/library/feedback/subagent x6). One waiver mechanism (`LintRegistryEntry.waiver`, architect-approved).
**OP-A rows to carry (cross-doc obligations):**
- **DOC20** -- render the Sec. G surfaces (G-07, G-08, G-10, G-11, G-12...G-16, G-20) as `DerivedReadModelRecord`-backed UI; matter-scoped (D-24).
- **Addenda A** -- full `NoVerdictReason` consolidation absorbing the R203 taxonomy (A-07); the canonical enum + alias registry land now, the merge when Addenda A next opens.
- **DOC12** -- `RoomKind.plan_review` registration (`OBL-DOC12-FORUM-01`, already exists; B-15).
- **DOC73** -- `LibraryPromotionPolicy` gate (NR.4 / D-20).
- **Review Studio** -- `OBL-RS-01...10` (already tracked in `DOC23_ADDB_Review_Studio.md`).
**Corrections baked in:** `confidence_merge_strategy` removed (was -> A-04, route-resolution); `tiebreaker_epsilon` removed (was -> B-16, a concurrency row); `A-02` not elevated (HIGH, not CRITICAL); `B-25` de-CRITICAL.
**Integrity:** Sec. E (held, Phase-B) and Sec. F (declined) intact and unmodified; the finding model is shared with (not forked from) the Review Studio unit; the `revision_in` chokepoint (A-16) is the single mutation path that the Review Studio human-assist also relies on.
---
## Status
**Card complete and self-contained (Layers 0-5 + new rows + R0.4 audit additions + Appendix R).** Every adjudicated row is now rendered in full (Appendix R) or in its governing Layer; the nine completeness-audit fold-ins, two new rows (NR.8-NR.9), and one DOC24 OP-A obligation are folded in. Ready for the multi-LLM red-team pass (ChatGPT 5 Pro + Gemini 2.5 Pro + Codex), then fold into the R0.4 spec edits.
**Awaiting your ruling (non-blocking):** the three parked items (`WorkspaceExternalizationPolicy`/`Receipt`; `ExternalSourceQueryPolicy`/`Receipt`; `TaskAgentForumSurfaceOwnership`) -- in or out of scope.
**Next deliverable still owed (separate):** the Test-set adjudication card (`DOC23_ADDENDA_B_TEST_SET_ADJUDICATION_CARD.md`) second-round reviews.