codex-local-rag builds a local hybrid search index over Codex history so you can ask things like:

• "Where did I debug that nginx reverse proxy issue?"
• "How did I fix a failing GitHub Actions workflow last month?"
• "What command did I use when I recovered that SQLite database?"

It is designed for real Codex artifacts, not generic chat logs. Instead of slicing everything into arbitrary chunks, it reconstructs turns, keeps tool output separate when it gets large, and combines sparse plus dense retrieval for better recall on both conceptual questions and exact technical strings.

Use this skill when the user asks for any of these:

Use this skill when the user asks for any of these:

Run the whole PR lifecycle without dropping the follow-through steps. Use this skill when the user wants a repeatable "open PR, wait, inspect review, fix, recheck, merge, verify deploy" loop for the current branch.

Last updated: March 10, 2026

This report documents why the existing Codex Desktop automation path for ragweld was treated as unsafe, what concrete failures we observed, what we changed locally to stop the lying/fake-green behavior, what we changed in the repo to harden the worker lanes, and what still remains open.

Primary trigger session:

Important: this is not about writing prettier logging copy. The point is to make operatorHint act like a tiny implementation diagnosis that has already done some of the debugging work for the next AI agent. The strongest mode is not supposed to produce vague junk like look at notifications. It is supposed to produce a compact root-cause note, more like PATCH /notifications -> parseNotificationPatchBody(); missing body.text guard before schema parse or RunPage reconnect effect opens a new EventSource before cleanup. In other words: route, function, boundary, missing guard, broken invariant, likely culprit. Not decorative logging.

These are real runs against the same tiny repo, same file, same error path.

Weak:

If this codebase is production, handles money, or touches sensitive data: treat this audit loop as a high-risk operation. Run with least privilege, avoid exporting long-lived credentials in your shell, and keep the agent in read-only mode.

For use with GitHub Copilot, Claude, or other AI coding assistants

Objective: Eliminate DOCX ↔ PDF round-trip layout drift through deterministic normalization.

Normalization Pipeline: