agentleFS
SonghongleiSonghonglei/better-agent-skills/skills/codex-thread-repair/SKILL.md

codex-thread-repair

Skill4 starsChanged 5 months agoMITView on GitHub

Diagnose and safely repair one local Codex Desktop task selected by its UI title or thread ID when recent turns are missing, truncated, merged into an interrupted turn, or hidden by damaged rollout ordering. Do not use for broad project/provider restoration or ordinary task continuation.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read references/repair-patterns.md when diagnosis is not healthy or supported_ordinal_orphan.

    line 36
  2. Distinctive:

    After Codex restarts, verify both layers:

    line 46
  3. Distinctive:

    Keep the backup until the user confirms the UI.

    line 54
  4. Very specific:

    Do not edit SQLite rows for this repair type.

    line 69

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Codex Thread Repair

Repair a specific local task without rewriting its message bodies or replacing the whole Codex state directory.

Dependencies

  • macOS with the Codex desktop app installed as /Applications/ChatGPT.app.
  • Python 3.10 or newer. The repair script uses only the Python standard library.
  • A local Codex state directory containing state_5.sqlite, sqlite/codex-dev.db, and the selected task's JSONL rollout. It defaults to $CODEX_HOME when set, otherwise ~/.codex.
  • Local read/write access to the selected rollout and Codex state. No network access or third-party Python package is required.

Workflow

  1. Resolve <skill-dir> as the directory containing this SKILL.md; never assume that the current working directory is the skill directory. Resolve and diagnose the user-supplied title or ID while Codex may remain open:

    python3 "<skill-dir>/scripts/thread_repair.py" diagnose "<title-or-id>"

    Exact UI-title matches come from local_thread_catalog; canonical rows and rollout paths come from state_5.sqlite. If a title is ambiguous, show the candidates and ask the user for an exact ID. Never choose by recency alone.

  2. Read references/repair-patterns.md when diagnosis is not healthy or supported_ordinal_orphan. Do not auto-repair an unrecognized structure.
  3. For supported_ordinal_orphan, explain the detected boundary and prepare one offline launcher:

    python3 "<skill-dir>/scripts/thread_repair.py" prepare "<title-or-id>"

    Tell the user to press Cmd+Q and then double-click the returned .command file. The launcher creates a targeted recovery set, applies the patch atomically, verifies it, and reopens Codex.

  4. After Codex restarts, verify both layers:

    python3 "<skill-dir>/scripts/thread_repair.py" verify "<title-or-id>"

    Also use the Codex task-reading tool to confirm that the formerly hidden recent turns are separate and readable.

  5. Keep the backup until the user confirms the UI. If rollback is required, have the user quit Codex and run:

    python3 "<skill-dir>/scripts/thread_repair.py" rollback "<backup-directory>" --reopen

Safety boundaries

  • Diagnosis and launcher preparation are read-only with respect to Codex state.
  • Mutation requires Codex Desktop, its renderer/services, and its app-server to be stopped.
  • Resolve the selected rollout to an absolute path under the selected Codex home's sessions/ directory. Refuse relative paths, symlink escapes, and database paths outside that directory.
  • Back up the target rollout plus consistent snapshots of state_5.sqlite, codex-dev.db, and .codex-global-state.json before replacement.
  • Preserve the original rollout byte-for-byte in the backup.
  • Auto-repair only the narrowly validated ordinal-regression/orphan-turn pattern. Refuse multiple regressions, ambiguous open turns, missing rollouts, malformed JSON, non-contiguous segments, active latest turns, or session-ID mismatches.
  • Preserve every original line except the top-level ordinal value after the repair boundary; insert exactly one turn_aborted event. Verify a normalized content hash before replacement.
  • Do not edit SQLite rows for this repair type. For provider mismatches, missing assignments, moved paths, or missing database rows, use recover-codex-project-chats instead.
  • On any post-replacement verification failure, restore the backed-up rollout automatically.

Commands

python3 "<skill-dir>/scripts/thread_repair.py" resolve "<title-or-id>"
python3 "<skill-dir>/scripts/thread_repair.py" diagnose "<title-or-id>"
python3 "<skill-dir>/scripts/thread_repair.py" prepare "<title-or-id>"
python3 "<skill-dir>/scripts/thread_repair.py" apply "<exact-thread-id>" --reopen
python3 "<skill-dir>/scripts/thread_repair.py" verify "<title-or-id>"
python3 "<skill-dir>/scripts/thread_repair.py" rollback "<backup-directory>" --reopen

--codex-home <path> is a global option and must appear before the subcommand. prepare --output <path> selects a new launcher path and refuses to overwrite an existing file. apply --reopen and rollback --reopen request reopening Codex; failure to reopen is reported as a warning without undoing a successful file operation.

Exit codes are 0 for success or no repair needed, 1 for an operational/safety error, 2 for an ambiguous selector or invalid CLI use, and 3 for an unsupported diagnosis or failed verification.

Report the resolved ID, UI title, rollout path, diagnosis, changed metadata, backup directory, integrity results, and UI verification outcome.

See CHANGELOG.md for release history.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.