Claude Skill

pitroom-implement

Use for well-defined code changes a cheaper worker can make - small bug fixes, simple features, mechanical refactors, renames, writing tests, fixing lint or type errors, docs - where you can state the goal and check the result with a diff or tests.

LLM Mart · 0 points · 2 views 0 listing impressions 0 install-command copies
Download ATASTECH-pitroom-skills_pitroom-implement-05f5f6a.zip · 1 KB

Install

skills CLI npx skills add https://github.com/ATASTECH/pitroom/tree/main/skills/pitroom-implement
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install atastech-pitroom@llmmart
Git git clone https://github.com/ATASTECH/pitroom.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole atastech/pitroom collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Implementing with a Pitroom worker

The worker edits code; you review and apply. Nothing reaches the user's tree without your decision.

Pick the mode

Flag Use
isolate -i Default for changes. The worker edits a private copy of the current state (uncommitted and untracked files included). You apply the patch.
write -w Only for trivial in-place edits while nobody else edits those files. One --write run per repository at a time; pitroom revert <run> undoes exactly its changes.

Run

pitroom run -i --link node_modules --verify "npm test -- auth" \
  "Fix: expired tokens return 500 instead of 401 in src/auth/session.ts. Add a regression test in test/auth.test.ts. Do not change the public API."
pitroom run -w "Rename getUserById to findUserById in src/ and update imports. No other changes."

--link shares ignored folders (node_modules, .venv) with the isolated copy so tests can run there; --verify runs your check after the worker and reports pass/fail; when it fails, its output first goes back to the worker for a fix round (--verify-fix N, default 1), and the report says passed after 1 fix round when that worked. Long changes: add --bg and collect with pitroom wait <run>. A change free models often get wrong: --race 3 gives it to three workers at once and keeps the best (the check decides first).

Write a good brief

  • Goal in one sentence, with the observable behaviour that must change.
  • Where: files, functions, the failing test or error message.
  • Constraints: "no new dependencies", "keep the public API", "match the existing style".
  • Done means: which tests must pass, what must not change.

Review and land

  1. Read the report: FILES CHANGED, VERIFICATION, OPEN ISSUES, and the verify: line.
  2. pitroom show <run> --patch: review the diff as you would a colleague's.
  3. isolate: pitroom apply <run> (checked apply; refuses on conflict, and on a patch that deletes files until you decide they are wanted: --allow-delete) or pitroom discard <run>. write: keep it, or pitroom revert <run>.
  4. Run the relevant tests yourself after applying.
  5. Needs another pass? pitroom run --continue <run> "Also handle the refresh-token path." (same session and copy; the patch accumulates).

Commit only if the user asked you to. If the worker fails twice, make the change yourself.

Files (pitroom)
  • SKILL.md 2.7 KB
    ---
    name: pitroom-implement
    description: Use for well-defined code changes a cheaper worker can make - small bug fixes, simple features, mechanical refactors, renames, writing tests, fixing lint or type errors, docs - where you can state the goal and check the result with a diff or tests.
    ---
    
    # Implementing with a Pitroom worker
    
    The worker edits code; you review and apply. Nothing reaches the user's tree without your decision.
    
    ## Pick the mode
    
    | | Flag | Use |
    |---|---|---|
    | isolate | `-i` | **Default for changes.** The worker edits a private copy of the current state (uncommitted and untracked files included). You apply the patch. |
    | write | `-w` | Only for trivial in-place edits while nobody else edits those files. One `--write` run per repository at a time; `pitroom revert <run>` undoes exactly its changes. |
    
    ## Run
    
    ```bash
    pitroom run -i --link node_modules --verify "npm test -- auth" \
      "Fix: expired tokens return 500 instead of 401 in src/auth/session.ts. Add a regression test in test/auth.test.ts. Do not change the public API."
    pitroom run -w "Rename getUserById to findUserById in src/ and update imports. No other changes."
    ```
    
    `--link` shares ignored folders (node_modules, .venv) with the isolated copy so tests can run there; `--verify` runs your check after the worker and reports pass/fail; when it fails, its output first goes back to the worker for a fix round (`--verify-fix N`, default 1), and the report says `passed after 1 fix round` when that worked. Long changes: add `--bg` and collect with `pitroom wait <run>`. A change free models often get wrong: `--race 3` gives it to three workers at once and keeps the best (the check decides first).
    
    ## Write a good brief
    
    - **Goal** in one sentence, with the observable behaviour that must change.
    - **Where**: files, functions, the failing test or error message.
    - **Constraints**: "no new dependencies", "keep the public API", "match the existing style".
    - **Done means**: which tests must pass, what must not change.
    
    ## Review and land
    
    1. Read the report: `FILES CHANGED`, `VERIFICATION`, `OPEN ISSUES`, and the `verify:` line.
    2. `pitroom show <run> --patch`: review the diff as you would a colleague's.
    3. isolate: `pitroom apply <run>` (checked apply; refuses on conflict, and on a patch that deletes files until you decide they are wanted: `--allow-delete`) or `pitroom discard <run>`. write: keep it, or `pitroom revert <run>`.
    4. Run the relevant tests yourself after applying.
    5. Needs another pass? `pitroom run --continue <run> "Also handle the refresh-token path."` (same session and copy; the patch accumulates).
    
    Commit only if the user asked you to. If the worker fails twice, make the change yourself.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.