agentleFS
memi-designmemi-design/design-skills/skills/codex-ops/SKILL.md

codex-ops

Skill7 starsChanged 4 months agoMITView on GitHub

Codex and Claude operating workflow for Memoire - JSON-first commands, commit hygiene, review loops, memory capture, and safe autonomous repo operation

The file

Show its rules and text

We split this skill into the rules it gives an agent

24 rules. About 0 are common, 19 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈19

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this when Codex or Claude is operating inside the Memoire repository and needs a stable workflow for safe autonomous changes.

    line 8
  2. Distinctive:

    Prefer machine-readable command output whenever Memoire exposes it.

    line 13
  3. Distinctive:

    Use dry-run modes before mutating files when available.

    line 30
  4. Distinctive:

    Verify with focused tests or typecheck.

    line 32
  5. Distinctive:

    Commit and push each lane separately.

    line 33
  6. Distinctive:

    Keep each commit scoped to one behavior or one workflow lane.

    line 37
  7. Distinctive:

    Do not mix note content, command changes, and unrelated generated output.

    line 38
  8. Distinctive:

    If the worktree is already dirty, treat existing changes as user state unless you know they were produced by your own current lane.

    line 39
  9. Distinctive:

    Do not rewrite or revert unrelated files to make your diff cleaner.

    line 41
  10. Distinctive:

    Read the diff before every commit.

    line 45
  11. Distinctive:

    Prefer targeted regression tests over broad reruns first.

    line 46
  12. Very specific:

    If a command has a --json mode, add or update tests against the structured payload rather than asserting on human-readable formatting.

    line 47
  13. Distinctive:

    When a fix changes behavior, verify both the focused path and the combined command suite that overlaps it.

    line 49
  14. Distinctive:

    If a workflow becomes stable, encode it in a built-in note or skill instead of repeating it in freeform prompts.

    line 54
  15. Very specific:

    If a command repeatedly needs structured output, add --json instead of teaching agents to parse ASCII tables or prose.

    line 56
  16. Distinctive:

    If a repository convention matters across many commits, capture it in a note, not only in one changelog entry or one conversation.

    line 58
  17. Distinctive:

    Prefer additive or explicitly scoped changes.

    line 63
  18. Very specific:

    Use --dry-run, doctor, and status --json before destructive or broad operations.

    line 64
  19. Very specific:

    Treat .memoire/, generated outputs, preview artifacts, and user specs as project state, not disposable scratch space.

    line 66
  20. Very specific:

    When a lane is already pushed to main, reconcile local work to HEAD instead of silently diverging.

    line 68
  21. Distinctive:

    add the markdown skill file

    line 87
  22. Distinctive:

    add a focused note regression test

    line 88
  23. Distinctive:

    run the note tests plus typecheck

    line 89
  24. Distinctive:

    commit and push that note as its own lane

    line 90

≈ “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 Ops

Use this when Codex or Claude is operating inside the Memoire repository and needs a stable workflow for safe autonomous changes.

Core Rule

Prefer machine-readable command output whenever Memoire exposes it. Do not scrape terminal prose if a JSON surface exists.

Current JSON-first commands:

  • memi status --json
  • memi connect --json
  • memi notes list --json
  • memi notes info <name> --json
  • memi compose "..." --json
  • memi doctor --json
  • memi generate --json
  • memi export --json

Execution Pattern

  1. Inspect state first with JSON-capable commands.
  2. Use dry-run modes before mutating files when available.
  3. Make one coherent change at a time.
  4. Verify with focused tests or typecheck.
  5. Commit and push each lane separately.

Commit Hygiene

  • Keep each commit scoped to one behavior or one workflow lane.
  • Do not mix note content, command changes, and unrelated generated output.
  • If the worktree is already dirty, treat existing changes as user state unless you know they were produced by your own current lane.
  • Do not rewrite or revert unrelated files to make your diff cleaner.

Review Loop

  • Read the diff before every commit.
  • Prefer targeted regression tests over broad reruns first.
  • If a command has a --json mode, add or update tests against the structured payload rather than asserting on human-readable formatting.
  • When a fix changes behavior, verify both the focused path and the combined command suite that overlaps it.

Memory Capture

  • If a workflow becomes stable, encode it in a built-in note or skill instead of repeating it in freeform prompts.
  • If a command repeatedly needs structured output, add --json instead of teaching agents to parse ASCII tables or prose.
  • If a repository convention matters across many commits, capture it in a note, not only in one changelog entry or one conversation.

Safe Autonomy

  • Prefer additive or explicitly scoped changes.
  • Use --dry-run, doctor, and status --json before destructive or broad operations.
  • Treat .memoire/, generated outputs, preview artifacts, and user specs as project state, not disposable scratch space.
  • When a lane is already pushed to main, reconcile local work to HEAD instead of silently diverging.

Recommended Sequence

For a normal implementation lane:

  1. memi status --json
  2. memi connect --json
  3. memi doctor --json
  4. edit the smallest viable set of files
  5. run focused tests
  6. run npm run typecheck
  7. commit one lane
  8. push immediately

For note work:

  1. add note.json
  2. add the markdown skill file
  3. add a focused note regression test
  4. run the note tests plus typecheck
  5. commit and push that note as its own lane

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.