agentleFS
Sign inStart free
SonghongleiSonghonglei/better-agent-skills/skills/recover-codex-project-chats/SKILL.md

recover-codex-project-chats

Skill4 starsChanged 5 months agoMITView on GitHub

Diagnose and safely repair Codex Desktop projects that exist but show “No chats/没有聊天”, while conversations may still appear under Recent. Use when users report missing project histories, lost thread-to-project mappings, model-provider changes, archived or invisible threads, moved cwd paths, incomplete ~/.codex restores, state_5.sqlite integrity or index problems, or need comparison against codex_threads_snapshot.csv or a ~/.codex backup.

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 4 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

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

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

    Run scripts/codex_project_recovery.sh diagnose [csv-path] while Codex may remain open.

    line 23
  2. Very specific:

    Read references/incident-patterns.md and classify the evidence before changing data.

    line 24
  3. Distinctive:

    Before any mutation, require Codex Desktop to be fully stopped.

    line 26
  4. Very specific:

    Run SQLite PRAGMA integrity_check on copies first.

    line 27
  5. Very specific:

    Run scripts/codex_project_recovery.sh verify while Codex is stopped, then restart Codex and check each project in the UI.

    line 33
  6. Very specific:

    Treat state_5.sqlite plus WAL/SHM, rollout JSONL, .codex-global-state.json, and project metadata as one recovery set.

    line 38
  7. Distinctive:

    Do not edit a live SQLite database.

    line 39
  8. Very specific:

    Do not replace all of ~/.codex when a targeted repair is sufficient.

    line 40
  9. Distinctive:

    Do not change provider metadata merely to make rows visible unless the configured provider is compatible with the original sessions.

    line 42
  10. Distinctive:

    Keep every pre-repair backup until the user verifies the UI and can open representative chats.

    line 43

≈ “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

Recover Codex project chats

Restore project visibility without overwriting the only copy of a conversation.

Workflow

  1. Run scripts/codex_project_recovery.sh diagnose [csv-path] while Codex may remain open. Save the output.
  2. Read references/incident-patterns.md and classify the evidence before changing data.
  3. Back up current ~/.codex, the supplied backup, and Codex-related Application Support directories. Exclude only live Unix sockets.
  4. Before any mutation, require Codex Desktop to be fully stopped. Confirm the app-server, renderer, and service processes are absent.
  5. Run SQLite PRAGMA integrity_check on copies first. Inspect .schema; never invent columns.
  6. Apply only the matching repair:
    • Provider mismatch: run scripts/codex_project_recovery.sh repair-provider <old-provider> <current-provider> after explaining the evidence. Repair only when both names refer to the same compatible backend.
    • Missing assignments: rebuild thread-project-assignments in .codex-global-state.json from actual thread IDs and cwd values. Preserve existing assignments. Resolve overlapping roots and temporary directories explicitly; do not rely solely on prefix matching.
    • Moved paths: create and review an old-cwd to current-project mapping before updating assignments or cwd metadata.
    • Missing database rows: merge only from a schema-compatible backup after integrity checks and a trial on copies.
  7. Run scripts/codex_project_recovery.sh verify while Codex is stopped, then restart Codex and check each project in the UI.
  8. Report root cause, backup paths, CSV rows, database rows, rollout files found, per-project visible counts, changed rows/files, and unrecovered thread IDs.

Safety rules

  • Treat state_5.sqlite plus WAL/SHM, rollout JSONL, .codex-global-state.json, and project metadata as one recovery set.
  • Do not edit a live SQLite database.
  • Do not replace all of ~/.codex when a targeted repair is sufficient.
  • Preserve archived state unless the user explicitly requests unarchiving.
  • Do not change provider metadata merely to make rows visible unless the configured provider is compatible with the original sessions. Prefer restoring the provider configuration when compatibility is uncertain.
  • Keep every pre-repair backup until the user verifies the UI and can open representative chats.

Resources

  • scripts/codex_project_recovery.sh: read-only diagnosis, guarded provider repair, and verification.
  • references/incident-patterns.md: known failure patterns, schema landmarks, and evidence tests.

Dependencies

  • macOS with Codex Desktop or ChatGPT Desktop using the Codex runtime
  • Zsh, SQLite 3, jq, ripgrep, rsync, and Perl
  • Optional CSV snapshot or ~/.codex backup for comparison

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.

Check your setup, free