feat(memory): warn before repeating rejected approaches

[ahrav]

Summary

Search now surfaces relevant rejected approaches as bounded warnings before an agent repeats them. Warnings remain advisory, reserve one result slot when relevant, and report retrieval, surfacing, missed-warning, and lifecycle counters.

This final PR completes the anti-memory stack after PR #75, PR #76, and PR #77.

Design

• Lexical and semantic retrieval share warn-only result semantics.
• One reserved top-K slot prevents a relevant warning from being crowded out.
• Rendering is bounded and strips warning text from positive memory surfaces.
• Verification archives stale records, expires old records, and preserves explicit retirement.
• Contract goldens cover plugin, Pi, and Rust tool schemas.

Validation

• bun run check:all
• 1,669 Rust tests passed; 48 skipped
• Focused warning, lifecycle, counter, tool-contract, and rendering tests

_{Stack created with GitHub Stacks CLI • Give Feedback 💬}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 06983d3d-7545-4d57-8f87-4f9e1294108a

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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

[chatgpt-codex-connector]

You have reached your Codex usage limits for security reviews. Please try again later.