Resolves #13967

Adds a focused proof fixture for the PR-review ingestion boundary. The mocked PR review cycle mirrors the #13915 shape: a CHANGES_REQUESTED review with Required Actions, a sanctioned [RETROSPECTIVE] note, and a later APPROVED review that neutralizes the blocker in prose. The test proves only the sanctioned tag becomes graph topology; raw review state, Required Actions, approvals, and rubber-stamps do not create routing weight.

Evidence: L2 (focused ingestion unit fixture proves current graph-topology boundary) -> L2 required (proof-first ticket; no runtime surface). No residuals.

Deltas from ticket

• Implements the proof as existing ingestion-spec coverage instead of adding parser topology.
• Leaves IssueIngestor.ingestPullRequestFeedback() behavior unchanged: only [KB_GAP], [TOOLING_GAP], [RETROSPECTIVE], and close-keyword issue links affect topology.
• Explicitly rejects raw PR_REVIEW, REQUIRED_ACTION, approval, blocker, request-changes, and neutralization edges as routing weight in the fixture.

Test Evidence

• git diff --check
• npm run agent-preflight -- test/playwright/unit/ai/services/ingestion/IssueIngestor.spec.mjs
• npm run test-unit -- test/playwright/unit/ai/services/ingestion/IssueIngestor.spec.mjs -> 10 passed.

Post-Merge Validation

• None; this is a unit-level proof of the ingestion boundary.

Commits

• 56358b4598 — test(ai): prove review actionability boundary (#13967)

Authored by Euclid (GPT-5, Codex Desktop). Session 019ef378-527d-7393-bc74-ec3a1d3f2ddf.