agentleFS
sidiangongyuansidiangongyuan/codex-skills-library/skills/codex-session-restore/SKILL.md

codex-session-restore

Skill11 starsChanged 4 months agoMITView on GitHub

Use when Codex Desktop sidebar sessions disappear, fail with systemError, or cannot continue after a model_provider switch. Diagnoses and repairs active, non-archived sessions without changing authentication or provider configuration.

The file

Show its rules and text

We split this skill into the rules it gives an agent

20 rules. About 0 are common, 4 are distinctive, and 16 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.

  • 16

    Very specific

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

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

    Use the bundled script first.

    line 10
  2. Very specific:

    If the user switched from krill/custom to official OAuth and wants old active sessions to appear under the current official login, use …

    line 18
  3. Very specific:

    If the user explicitly wants to switch the environment back to krill and make sessions belong to krill, use --provider krill.

    line 20
  4. Very specific:

    Do not edit config.toml, auth.json, or cc-switch's current provider just to restore sidebar visibility.

    line 22
  5. Very specific:

    Use repair-active --provider PROVIDER --apply --yes only after a dry run confirms the active non-archived session set is correct.

    line 49
  6. Very specific:

    Run scan-thread --thread-id ID before single-thread repair.

    line 62
  7. Very specific:

    Run the same command with --apply only after the preview is acceptable.

    line 67
  8. Very specific:

    Run verify --provider PROVIDER or verify --thread-id ID --provider PROVIDER.

    line 68
  9. Distinctive:

    If available, search the live Codex thread list for a known old title to confirm local sessions reappeared.

    line 69
  10. Very specific:

    When the user says, "I am logged in with official Codex now, but my krill sessions disappeared," keep official OAuth as-is and migrate …

    line 75
  11. Very specific:

    When the user explicitly wants to use krill as the current provider, repair active session metadata to krill:

    line 86
  12. Very specific:

    Do not touch archived_sessions or archived rows unless the user explicitly asks and the command includes --include-archived.

    line 97
  13. Distinctive:

    Do not switch provider configuration or copy third-party auth settings unless the user explicitly asks to change providers.

    line 99
  14. Distinctive:

    Do not edit live files manually when the script can perform the same repair.

    line 102
  15. Very specific:

    Do not use task_complete to close an interrupted or damaged turn.

    line 103
  16. Very specific:

    Do not delete non-empty historical turns.

    line 105
  17. Very specific:

    Do not close recent open turns unless the user explicitly confirms they are stale.

    line 107
  18. Distinctive:

    Prefer single-thread repair before bulk repair unless the user explicitly asks to restore all active sessions.

    line 111
  19. Very specific:

    Update threads.model_provider and rollout session_meta provider metadata.

    line 116
  20. Very specific:

    Read references/recovery-checklist.md when deciding whether to repair provider metadata only, repair rollout structure, close child edges, …

    line 130

≈ “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 Session Restore

Use the bundled script first. It is dry-run by default and writes nothing unless --apply is present.

Provider Direction Rule

Set --provider to the provider that Codex Desktop should recognize now, not automatically to the provider that originally created the sessions.

  • If the user switched from krill/custom to official OAuth and wants old active sessions to appear under the current official login, use --provider openai.
  • If the user explicitly wants to switch the environment back to krill and make sessions belong to krill, use --provider krill.
  • Do not edit config.toml, auth.json, or cc-switch's current provider just to restore sidebar visibility. Changing current auth/provider is a separate user request.

Quick Start

Resolve <skill-dir> to the directory containing this SKILL.md; do not assume a particular global skills installation root.

Restore old krill sessions into the current official/OAuth sidebar:

python "<skill-dir>/scripts/restore_codex_sessions.py" scan --provider openai
python "<skill-dir>/scripts/restore_codex_sessions.py" repair-active --provider openai
python "<skill-dir>/scripts/restore_codex_sessions.py" repair-active --provider openai --apply --yes
python "<skill-dir>/scripts/restore_codex_sessions.py" verify --provider openai

Repair one thread:

python "<skill-dir>/scripts/restore_codex_sessions.py" scan-thread --thread-id THREAD_ID --provider openai
python "<skill-dir>/scripts/restore_codex_sessions.py" repair-thread --thread-id THREAD_ID --provider openai
python "<skill-dir>/scripts/restore_codex_sessions.py" repair-thread --thread-id THREAD_ID --provider openai --apply
python "<skill-dir>/scripts/restore_codex_sessions.py" verify --thread-id THREAD_ID --provider openai

Use repair-active --provider PROVIDER --apply --yes only after a dry run confirms the active non-archived session set is correct. Open turns newer than 30 minutes are treated as possibly active and are not closed unless --close-recent-open-turns is explicitly present. Archived sessions stay untouched unless --include-archived is explicitly present.

Workflow

  1. Identify the target thread id or run scan to find active non-archived threads with provider mismatch, parse errors, open turns, open child spawn edges, or stale goals.
  2. Decide the target provider from the user's current login/config goal: current official OAuth means openai; current krill means krill.
  3. Run scan-thread --thread-id ID before single-thread repair. Read references/recovery-checklist.md when the report shows multiple issue types or Codex Desktop still shows systemError.
  4. Run repair-thread --thread-id ID --provider PROVIDER or repair-active --provider PROVIDER without --apply to preview actions.
  5. Run the same command with --apply only after the preview is acceptable.
  6. Run verify --provider PROVIDER or verify --thread-id ID --provider PROVIDER.
  7. If available, search the live Codex thread list for a known old title to confirm local sessions reappeared. Otherwise fully restart Codex Desktop so the app server reloads thread state from disk.

Common Provider Switch Recipes

When the user says, "I am logged in with official Codex now, but my krill sessions disappeared," keep official OAuth as-is and migrate active session metadata to openai:

python "<skill-dir>/scripts/restore_codex_sessions.py" scan --provider openai
python "<skill-dir>/scripts/restore_codex_sessions.py" repair-active --provider openai
python "<skill-dir>/scripts/restore_codex_sessions.py" repair-active --provider openai --apply --yes
python "<skill-dir>/scripts/restore_codex_sessions.py" verify --provider openai

When the user explicitly wants to use krill as the current provider, repair active session metadata to krill:

python "<skill-dir>/scripts/restore_codex_sessions.py" repair-active --provider krill
python "<skill-dir>/scripts/restore_codex_sessions.py" repair-active --provider krill --apply --yes
python "<skill-dir>/scripts/restore_codex_sessions.py" verify --provider krill

Safety Rules

  • Do not touch archived_sessions or archived rows unless the user explicitly asks and the command includes --include-archived.
  • Do not switch provider configuration or copy third-party auth settings unless the user explicitly asks to change providers. Session restore normally repairs metadata to match the provider already in use.
  • Do not edit live files manually when the script can perform the same repair.
  • Do not use task_complete to close an interrupted or damaged turn. Use turn_aborted.
  • Do not delete non-empty historical turns. The script isolates open turns by inserting or appending turn_aborted.
  • Do not close recent open turns unless the user explicitly confirms they are stale. Use --close-recent-open-turns only for known failed continuations.
  • Back up before every write. The script creates session-restore-backup-YYYYMMDD-HHMMSS under CODEX_HOME.
  • Prefer single-thread repair before bulk repair unless the user explicitly asks to restore all active sessions.

Script Capabilities

  • Update threads.model_provider and rollout session_meta provider metadata.
  • Detect rollout parse errors, missing rollout files, open task turns, and provider metadata drift.
  • Coalesce duplicate turn_context entries by unique turn_id; ignore legacy no-turn_id contexts for mutation.
  • Insert one turn_aborted event per stale open turn.
  • Close stale thread_spawn_edges after making child rollouts internally closed.
  • Report non-complete thread_goals; optionally complete them only when an explicit repair flag is used.
  • Print provider distributions, backup paths, modified files, and restart guidance.

Reference

Read references/recovery-checklist.md when deciding whether to repair provider metadata only, repair rollout structure, close child edges, complete stale goals, or ask the user to restart Codex Desktop.

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.