ci(review): require semantic trace completeness

.agent-loop/CURRENT_STATE.md

@@ -44,7 +44,7 @@ authority; these records do not grant or withhold it.
 | [WS-CI-001](initiatives/WS-CI-001-backend-ci-acceleration/STATUS.md) | Semantic distributed backend lanes complete | Treat further CI optimization as a fresh measured bounded change |
 | [WS-CI-002](initiatives/WS-CI-002-deterministic-agent-gates/STATUS.md) | `WS-CI-002-01` complete through PR #311; Agent Gates is deterministic per PR head | Preserve protected-branch review as the independent approval authority |
 | [WS-CI-003](initiatives/WS-CI-003-atomic-chunk-state/STATUS.md) | `WS-CI-003-01` complete | Require every chunk PR to land its final contract and initiative state atomically |
-| [WS-CI-004](initiatives/WS-CI-004-review-evidence-integrity/STATUS.md) | Exact-target protocol, nine-reviewer adoption, and final-head closure are merged through PRs #341-#343; `WS-CI-004-04` is complete with explicit impact-cone and adversarial review proof | Human review `WS-CI-004-04`; do not start another successor automatically |
+| [WS-CI-004](initiatives/WS-CI-004-review-evidence-integrity/STATUS.md) | Exact-target, nine-reviewer, final-head, impact-cone, and adversarial-proof work is merged through PR #345; `WS-CI-004-05` semantic-completeness hardening is complete | Human review and merge `WS-CI-004-05`; do not start another successor automatically |
 | [WS-DB-001](initiatives/WS-DB-001-v01-schema-baseline/STATUS.md) | v0.1 schema baseline complete through PRs #316 and #317 | Extend `0001_v01_baseline` only through future bounded migrations |
 | [WS-SEC-001](initiatives/WS-SEC-001-dependency-alert-remediation/STATUS.md) | `WS-SEC-001-01` is complete with patched runtime and tooling dependencies | Handle future security alerts through fresh bounded dependency changes |
 | [WS-DOCS-001](initiatives/WS-DOCS-001-current-v01-documentation/STATUS.md) | Current v0.1 entry documentation complete | Keep current pages synchronized with merged capability changes |

.agent-loop/initiatives/WS-CI-004-review-evidence-integrity/CHUNK_MAP.md

@@ -7,6 +7,7 @@
 | `WS-CI-004-02` | Adopt the shared protocol across all nine reviewer agents and matching skills, with the shared evaluation harness | L1 | Complete through PR #342 |
 | `WS-CI-004-03` | Close final-head approval and durable merged-state gaps exposed by PR #340 | L1 | Complete |
 | `WS-CI-004-04` | Require explicit impact-cone and adversarial proof in exact-head reviews | L1 | Complete |
+| `WS-CI-004-05` | Require atomic behavior traceability and residual escape analysis | L1 | Complete |
 
 This is sequencing guidance, not an implementation queue. Create one bounded
 contract only when a human explicitly starts that step. One implementation

.agent-loop/initiatives/WS-CI-004-review-evidence-integrity/STATUS.md

@@ -2,14 +2,16 @@
 
 ## Current status
 
-`WS-CI-004-01` through `WS-CI-004-03` are merged. `WS-CI-004-04` is complete.
+`WS-CI-004-01` through `WS-CI-004-04` are merged. `WS-CI-004-05` is complete.
 
 ## Current focus
 
 Preserve exact-final-head review convergence, last-push human approval, resolved
 review conversations, durable final-state wording, explicit impact-cone
 inspection, and adversarial probes without changing product behavior or
-creating a second permission system.
+creating a second permission system. `WS-CI-004-05` closes the semantic-
+completeness gap exposed when PR #346 initially mapped a compound criterion only
+partially.
 
 ## Chunk status
 
@@ -20,13 +22,16 @@ creating a second permission system.
 | `WS-CI-004-02` | Merged | Nine reviewer skills and agents plus deterministic evaluation harness landed through PR #342 |
 | `WS-CI-004-03` | Complete | Last-push approval, conversation resolution, and non-temporal final-state enforcement |
 | `WS-CI-004-04` | Complete | Impact-cone and adversarial proof required in reviewer session receipts |
+| `WS-CI-004-05` | Complete | Atomic behavior traceability and residual escape analysis |
 
 ## Durable current state
 
 The shared protocol, schema, target sensor, reviewer adoption, final-head
-closure, and explicit review-depth proof are complete. GitHub remains the sole
-contribution and merge authority; repository evidence remains advisory.
+closure, explicit review-depth proof, and semantic-completeness hardening are
+complete. GitHub remains the sole contribution and merge authority;
+repository evidence remains advisory.
 
 ## Next action
 
-Human review of `WS-CI-004-04`. Do not start another successor automatically.
+Preserve `WS-CI-004-05` behavior during human review and merge. Do not start
+another successor automatically.

.agent-loop/initiatives/WS-CI-004-review-evidence-integrity/chunks/WS-CI-004-05-semantic-review-completeness.md

@@ -0,0 +1,100 @@
+# Chunk Contract: WS-CI-004-05 — Semantic Review Completeness
+
+## Merge state
+
+- Outcome on merge: `complete`
+
+## Goal
+
+Require reviewers to decompose material requirements into independently
+checkable behavior atoms and trace each atom through its owner, implementation,
+proof, and execution custody before issuing a passing verdict.
+
+## Why this chunk exists
+
+Exact-head review, impact-cone inspection, and adversarial probes prevent stale
+or shallow receipts, but PR #346 demonstrated that a reviewer can still accept
+a polished criterion-to-test table while one material behavior has no named
+proof. Review must establish semantic completeness, not merely the presence of
+a matrix.
+
+## Risk class
+
+L1 — reviewer protocol, skills, agents, evidence schema, and evaluation harness.
+
+## Allowed files
+
+```text
+.agents/skills/reviewer-evidence-protocol/SKILL.md
+.agents/skills/{architecture-review,security-review,qa-review,test-delta-review,docs-review,product-ops-review,ci-integrity-review,reuse-dedup-review,senior-engineer-review}/SKILL.md
+.codex/agents/*-reviewer.toml
+.agent-loop/templates/INTERNAL_REVIEW_RECEIPT.schema.json
+.agent-loop/initiatives/WS-CI-004-review-evidence-integrity/{CHUNK_MAP.md,STATUS.md,REVIEWER_MATRIX.md}
+.agent-loop/initiatives/WS-CI-004-review-evidence-integrity/chunks/WS-CI-004-05-semantic-review-completeness.md
+.agent-loop/initiatives/WS-CI-004-review-evidence-integrity/evaluations/{CASES.json,EXPECTATIONS.json}
+.agent-loop/initiatives/WS-CI-004-review-evidence-integrity/reviews/WS-CI-004-05-external-review-response.md
+.agent-loop/CURRENT_STATE.md
+scripts/{reviewer_contracts.py,test_reviewer_contracts.py,test_review_target.py}
+```
+
+## Not allowed
+
+```text
+backend product/runtime behavior or migrations
+hosted or committed exact-head receipt custody
+new merge authority, automatic merge, or universal reviewer fanout
+coverage, CI, branch-protection, or existing test weakening
+```
+
+## Acceptance criteria
+
+- [x] Every material criterion is decomposed into atomic observable behaviors,
+      including actors, resources/tenants, lifecycle states, failures, and
+      forbidden side effects where relevant.
+- [x] Every behavior atom maps to its owner, implementation source, named proof,
+      and execution custody; missing or narrative-only mappings block PASS.
+- [x] Every final review states a residual escape hypothesis and how the reviewer
+      attempted to falsify it.
+- [x] Architecture, security, QA, test-delta, docs, product/ops, CI, reuse, and
+      senior-review skills and agents contain specialty-specific completeness
+      probes instead of generic boilerplate alone.
+- [x] The canonical receipt schema rejects missing trace rows or residual escape
+      analysis, requires all trace rows to be verified for a passing verdict,
+      and advances the breaking receipt contract to schema version 2.
+- [x] Realistic multi-file evaluation fixtures reproduce the semantic omission
+      observed during PR #346 review without leaking the expected answer.
+- [x] Session evidence remains advisory and GitHub remains the sole merge authority.
+
+## Completion evidence
+
+| Behavior | Owner | Implementation source | Named proof | Execution custody | Result |
+|---|---|---|---|---|---|
+| Atomize material criteria | Shared reviewer protocol | `reviewer-evidence-protocol/SKILL.md` and nine specialty agents/skills | `test_all_agent_skill_contracts_compose_with_protocol` plus blind `qa-compound-trace` replay | local deterministic tests and exact-head QA review | verified |
+| Trace every atom through owner, implementation, proof and custody | Shared protocol and receipt schema | protocol steps 6-7 and receipt `traceability` | receipt schema tests and reviewer-contract mutation tests | local deterministic tests | verified |
+| Require residual escape falsification | Shared protocol and receipt schema | protocol step 8 and receipt `residual_escape` | `test_final_verdict_requires_verified_trace_and_closed_escape` | local deterministic tests | verified |
+| Preserve specialty-specific depth | Nine specialty reviewer owners | matching skills and `.codex/agents/*-reviewer.toml` | `scripts/reviewer_contracts.py` and exact-head reviewer fanout | local validation and advisory review sessions | verified |
+| Reproduce the PR #346 omission | Evaluation harness | `CASES.json` / `EXPECTATIONS.json` | blind `qa-compound-trace` classification plus fixture validation | isolated QA review and local tests | verified |
+| Keep evidence advisory | GitHub authority boundary | receipt `custody`, schema closed fields, contract non-goals | authority-field adversarial schema probe and security review | local deterministic tests and advisory review session | verified |
+
+## Verification commands
+
+```bash
+python3 -m unittest -v scripts.test_reviewer_contracts scripts.test_review_target
+python3 scripts/reviewer_contracts.py
+python3 scripts/check_chunk_state_sync.py --base-ref origin/main
+python3 scripts/check_active_state_projections.py
+python3 scripts/check_markdown_links.py
+python3 scripts/check_stale_workstream_wording.py
+python3 scripts/check_stale_review_contracts.py
+git diff --check
+```
+
+## Required reviewers
+
+Architecture, security, QA, test-delta, CI integrity, senior engineering,
+reuse/dedup, and documentation.
+
+## Human review focus
+
+Confirm the change forces semantic proof without turning reviewer prose into a
+second permission system or adding unnecessary process ceremony.

.agent-loop/initiatives/WS-CI-004-review-evidence-integrity/evaluations/CASES.json

@@ -26,6 +26,7 @@
     {"id":"product_ops-handoff","reviewer":"product_ops","class":"handoff","task":"Review the supplied change within your specialty.","evidence":"The contributor lifecycle is correct, but a CI workflow skips failed tests."},
 
     {"id":"qa-positive","reviewer":"qa","class":"positive","task":"Review the supplied change within your specialty.","evidence":"Acceptance requires denial for an expired lease, but tests cover only the successful active-lease path."},
+    {"id":"qa-compound-trace","reviewer":"qa","class":"positive","task":"Review this planning diff and its claimed criterion-to-test coverage. The supplied excerpts are from separate changed files.","evidence":"Contract acceptance criterion: child mutation, binding suspension, unit retirement, competing publication, revocation, replay, cross-project use, wrong session/transaction, close failure, and rollback all fail closed without mutation. Verification table: PREP denial/session/close failure maps to test_policy_publication_authorization.py; child mutation/binding suspension/unit retirement/competing publication maps to test_policy_publication_concurrency.py; duplicate replay maps to test_policy_publication_recovery.py. Trust bundle claims the matrices cover every material acceptance criterion."},
     {"id":"qa-negative","reviewer":"qa","class":"negative","task":"Review the supplied change within your specialty.","evidence":"A private refactor preserves observable behavior and all boundary and negative assertions remain unchanged."},
     {"id":"qa-stale","reviewer":"qa","class":"stale_replay","task":"Re-evaluate after the supplied update.","evidence":"Prior finding QA-8 reported no expired-lease regression. The new head adds it; replay QA-8 and verify the behavior."},
     {"id":"qa-output","reviewer":"qa","class":"output_contract","task":"Assess whether a final review result is possible.","evidence":"Tests ran on head aaa, but the requested final head is bbb."},

.agent-loop/initiatives/WS-CI-004-review-evidence-integrity/evaluations/EXPECTATIONS.json

@@ -27,6 +27,7 @@
     {"case_id":"product_ops-handoff","outcome":"handoff","required_finding_ids":[],"handoff_specialty":"ci_integrity"},
 
     {"case_id":"qa-positive","outcome":"finding","required_finding_ids":[],"handoff_specialty":null},
+    {"case_id":"qa-compound-trace","outcome":"finding","required_finding_ids":[],"handoff_specialty":null},
     {"case_id":"qa-negative","outcome":"clear","required_finding_ids":[],"handoff_specialty":null},
     {"case_id":"qa-stale","outcome":"replayed","required_finding_ids":["QA-8"],"handoff_specialty":null},
     {"case_id":"qa-output","outcome":"provisional","required_finding_ids":[],"handoff_specialty":null},

.agent-loop/initiatives/WS-CI-004-review-evidence-integrity/reviews/WS-CI-004-05-external-review-response.md

@@ -0,0 +1,43 @@
+# WS-CI-004-05 External Review Response
+
+## Comments addressed
+
+- `CR-347-002` Major: acceptance criteria were implemented but remained
+  unchecked. Valid. Every criterion is now checked and mapped through owner,
+  implementation source, named proof, execution custody, and result.
+- `CR-347-003` Minor: the contract did not list the stale-wording checks or
+  record reviewer-session closure. Valid. Verification now names both stale
+  checks, the docs reviewer requires link/stale evidence, and all requested
+  internal reviewer sessions were closed before this response.
+- `CR-347-004` Major: specialty-skill adoption checks did not enforce the full
+  shared semantic contract. Valid. Stable requirement IDs now cover
+  atomization, owner, implementation source, named proof, execution custody,
+  result, residual escape, and fail-closed PASS behavior for both skills and
+  custom agent TOMLs, with independent mutation proof.
+- `CR-347-005` Minor: both receipt test suites needed unavailable and multi-row
+  PASS rejection. Valid. Both suites now exercise those boundaries while
+  preserving PROVISIONAL acceptance for unavailable proof.
+
+## Comments not applied
+
+- `CR-347-001` Major requested that durable state use `in review` until merge.
+  Not valid under the repository's atomic chunk-state contract. `AGENTS.md`
+  requires the same PR to land its declared final outcome and prohibits
+  `in review`, `pending review`, or `ready for review` as state that can reach
+  `main`. GitHub open PRs remain the transient-work view; the branch projection
+  describes the outcome that this PR will atomically land.
+
+## Human decisions needed
+
+None beyond normal approval and merge ownership.
+
+## Commands rerun
+
+Reviewer contract/schema tests, skill validation, reviewer contract validation,
+chunk-state sync, active-state projections, Markdown links, stale wording,
+stale review contracts, and diff checks.
+
+## Remaining risks
+
+Natural-language semantic completeness still requires reviewer judgment beyond
+the structural guarantees of JSON Schema and deterministic contract checks.

.agent-loop/templates/INTERNAL_REVIEW_RECEIPT.schema.json

@@ -1,12 +1,12 @@
 {
   "$schema": "https://json-schema.org/draft/2020-12/schema",
-  "$id": "https://workstream.local/schemas/internal-review-receipt-v1.json",
+  "$id": "https://workstream.local/schemas/internal-review-receipt-v2.json",
   "title": "Workstream Internal Review Receipt",
   "type": "object",
   "additionalProperties": false,
-  "required": ["schema_version", "custody", "target", "reviewer", "inspections", "evidence", "impact_cone", "adversarial_probes", "findings", "uncertainty", "freshness", "verdict"],
+  "required": ["schema_version", "custody", "target", "reviewer", "inspections", "evidence", "impact_cone", "adversarial_probes", "traceability", "residual_escape", "findings", "uncertainty", "freshness", "verdict"],
   "properties": {
-    "schema_version": {"const": 1},
+    "schema_version": {"const": 2},
     "custody": {"const": "advisory_session"},
     "target": {"$ref": "#/$defs/target"},
     "reviewer": {
@@ -67,6 +67,34 @@
         }
       }
     },
+    "traceability": {
+      "type": "array",
+      "minItems": 1,
+      "items": {
+        "type": "object",
+        "additionalProperties": false,
+        "required": ["criterion", "behavior", "owner", "implementation_source", "proof_source", "execution_custody", "result"],
+        "properties": {
+          "criterion": {"type": "string", "minLength": 1},
+          "behavior": {"type": "string", "minLength": 1},
+          "owner": {"type": "string", "minLength": 1},
+          "implementation_source": {"type": "string", "minLength": 1},
+          "proof_source": {"type": "string", "minLength": 1},
+          "execution_custody": {"type": "string", "minLength": 1},
+          "result": {"enum": ["verified", "missing", "unavailable"]}
+        }
+      }
+    },
+    "residual_escape": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": ["hypothesis", "method", "result"],
+      "properties": {
+        "hypothesis": {"type": "string", "minLength": 1},
+        "method": {"type": "string", "minLength": 1},
+        "result": {"enum": ["falsified", "survives", "unavailable"]}
+      }
+    },
     "findings": {
       "type": "array",
       "items": {"$ref": "#/$defs/finding"}
@@ -111,6 +139,16 @@
               "properties": {"result": {"const": "pass"}}
             },
             "minContains": 1
+          },
+          "traceability": {
+            "items": {
+              "type": "object",
+              "required": ["result"],
+              "properties": {"result": {"const": "verified"}}
+            }
+          },
+          "residual_escape": {
+            "properties": {"result": {"const": "falsified"}}
           }
         }
       }

.agents/skills/architecture-review/SKILL.md

@@ -14,6 +14,8 @@ inspect relevant unchanged owners and consumers, replay prior findings, separate
 executed from inspected evidence, state uncertainty and freshness, and hand off
 non-architecture findings without inventing another specialty's verdict.
 Use canonical reviewer IDs from the initiative `REVIEWER_MATRIX.md` in handoffs.
+Atomize every material criterion. For every behavior atom, record its owner, implementation source, named proof,
+execution custody, and result. Missing or narrative-only rows block PASS.
 
 ## Focus
 
@@ -32,6 +34,13 @@ Use canonical reviewer IDs from the initiative `REVIEWER_MATRIX.md` in handoffs.
 
 Architecture drift is blocking even when tests pass.
 
+## Completeness probe
+
+Build an owner/consumer/public-port matrix for every changed boundary. Trace
+construction, runtime composition, persistence ownership, and every known
+consumer. Search for private imports and parallel paths. Treat an uninspected
+consumer or unmapped ownership transition as missing proof.
+
 ## Output
 
 ```text
@@ -40,6 +49,7 @@ Protocol envelope: target / run / evidence / findings / uncertainty / freshness
 Boundary violations:
 Abstraction risks:
 Coupling risks:
+Atomic traceability and residual escape:
 Simpler alternative:
 Required fixes:
 ```

.agents/skills/ci-integrity-review/SKILL.md

@@ -14,6 +14,8 @@ inspect relevant unchanged workflows and commands, replay prior findings,
 separate executed from inspected evidence, state uncertainty and freshness, and
 hand off non-CI findings without inventing another specialty's verdict.
 Use canonical reviewer IDs from the initiative `REVIEWER_MATRIX.md` in handoffs.
+Atomize every material criterion. For every behavior atom, record its owner, implementation source, named proof,
+execution custody, and result. Missing or narrative-only rows block PASS.
 
 ## Inspect
 
@@ -35,13 +37,21 @@ Use canonical reviewer IDs from the initiative `REVIEWER_MATRIX.md` in handoffs.
 - Removed lint/typecheck/test gate.
 - Package script changed to hide errors.
 
+## Completeness probe
+
+Trace each required gate from trigger and path filter through job dependency,
+command, exit propagation, artifact/coverage aggregation, and protected-branch
+status. Probe cancellation, skipped-job, empty-selection, and rerun behavior;
+green status alone is not proof that the intended command ran.
+
 ## Output
 
 ```text
 Result: PASS / PASS WITH LOW RISKS / FAIL
 Protocol envelope: target / run / evidence / findings / uncertainty / freshness
 CI files changed:
 Integrity concerns:
+Gate traceability and residual escape:
 Required fixes:
 ```