docs(rag): A3 Phase-2 gate check — findings #12–14 remain deferred

[BigSimmo]

What & why

Task A3 "RAG optimization Phase 2 (architectural)" re-examined injection-threat findings #12–14 (source-authority tier, per-claim prose entailment, structural role separation) against the standing rule:

No re-index / retrieval change without a real golden miss proving a gain — and eval:retrieval:quality must improve, not merely not-regress.

Outcome: all three stay correctly deferred. The gate is not met, so no code change was made. This PR records the disposition durably in docs/rag-injection-threat-model.md §8 so the decision (and the precise unblock conditions) don't get re-litigated next session.

Disposition (per finding)

• No concrete golden-eval miss exists. Retrieval golden set is all-green on record (document_recall@5=1.0, content_recall@5=1.0, top_k_hit_rate=1.0; reindex gate content_mrr@10≈0.93). rag_query_misses → rag_aliases promotion is still privacy-blocked, so nothing has been promoted to explore against.
• [codex] Clinical KB cleanup readiness #13 / Clean stale frontend identity and CSS #14 don't touch retrieval ranking, so eval:retrieval:quality can't register a gain from them — they're answer-path injection mitigations that belong to the seeded-chunk injection harness (mitigation Improve Clinical KB dashboard and RAG hardening #11, live-model-gated), not a retrieval-gated task.
• Align project dependencies and cleanup stale branches #12 (source-authority tier) is the only retrieval-touching one, and on an all-green set the realistic outcome is a recall regression — reintroducing authority as a primary sort key is what PR RAG optimisation: answer quality (P1-P10), v17 numeric grounding, offline fallback, hybrid RPC hardening + golden-recall regression fix #118 removed. Stays deferred until a promoted miss shows authority-as-tie-breaker lifts eval:retrieval:quality.

Scope / safety

• Docs-only. No code, retrieval, ranking, or answer-path change.
• eval:retrieval:quality is provider-backed (scripts/eval-retrieval.ts → requireServerEnv + requireOpenAIEnv, live Supabase + OpenAI) and was not run — speculative miss-hunting is owner-approved spend.
• Clinical governance preflight: N/A — no ingestion, answer-generation, retrieval, ranking, source-rendering, privacy, or production-config code is touched.

Verification

• prettier --write docs/rag-injection-threat-model.md → unchanged (already conforms to the format gate).

🤖 Generated with Claude Code

[supabase]

This pull request has been ignored for the connected project sjrfecxgysukkwxsowpy because there are no changes detected in supabase directory. You can change this behaviour in Project Integrations Settings ↗︎.

---

Preview Branches by Supabase.
Learn more about Supabase Branching ↗︎.

[coderabbitai]

Warning

Review limit reached

@BigSimmo, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 34 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 204a314d-1e4b-4f49-967c-6fb2ec691209

📥 Commits

Reviewing files that changed from the base of the PR and between 33edba1 and 501b949.

📒 Files selected for processing (1)

• docs/rag-injection-threat-model.md

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/rag-optimization-phase-2-748178

✨ Simplify code

• Create PR with simplified code
• Commit simplified code in branch claude/rag-optimization-phase-2-748178

---

_{Comment @coderabbitai help to get the list of available commands.}