agentleFS
HazATHazAT/pi-config/skills/worker/SKILL.md

worker

Skill451 starsChanged 7 months agoMITView on GitHub

Implement one todo or review-fix task supplied by a coordinator, verify it, create one focused commit, and report evidence through the current Superconductor session. Use when launched by /execute, launched from a review repair loop, or asked to implement one assigned task.

We split this skill into the rules it gives an agent

6 rules. 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 8 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the supplied handover files directly and treat them as read-only.

    line 19
  2. Distinctive:

    Keep changes limited to the assigned task and preserve unrelated work.

    line 23
  3. Distinctive:

    If verification fails, do not commit partial work.

    line 27
  4. Very specific:

    Read and follow ~/.pi/agent/skills/commit/SKILL.md.

    line 33
  5. Distinctive:

    Create exactly one focused commit.

    line 35
  6. Distinctive:

    Verify it with git show --stat --oneline HEAD and capture the full SHA.

    line 36

“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “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

Worker

Implement exactly the assigned todo or review-fix task. Do not redesign the plan, expand scope, launch agents, push, or edit coordinator handover files.

Inputs

The launch prompt must provide:

  • available absolute plan, todos, and review paths;
  • one task ID and its exact scope;
  • the checkout to use;
  • any additional acceptance constraints.

Read the supplied handover files directly and treat them as read-only. Inspect repository instructions, git status --short, and every target file before editing. If context is missing or unrelated dirty changes make an isolated commit unsafe, stop and report a blocker rather than guessing.

Implement and Verify

Keep changes limited to the assigned task and preserve unrelated work. Run the smallest meaningful tests, typecheck, build, or smoke checks required by its acceptance criteria. Remove temporary/debugging artifacts.

Run every mutation-capable verification command before the commit. After the final verification, inspect git status --short: include legitimate generated source changes in scope, or restore disposable generated output before reporting. If a post-commit hook or required check changes tracked files, reconcile those files, rerun the affected verification, and amend or create the requested single focused commit. Never report completion with task-owned changes left dirty.

If verification fails, do not commit partial work. Report the command, failure, and blocker.

Commit

After verification passes:

  1. Read and follow ~/.pi/agent/skills/commit/SKILL.md.
  2. Review the complete diff and stage only this task's files.
  3. Create exactly one focused commit; do not push.
  4. Verify it with git show --stat --oneline HEAD and capture the full SHA.
  5. Confirm git status --short contains no uncommitted changes owned by this task.

A no-op task may omit a commit only when the evidence proves no source change was needed.

Final Report

Return a concise final response for the coordinator to collect with sc agent read:

Status: DONE | BLOCKED
Task: TODO-NNN | REVIEW-FIX-NNN
Summary: ...
Files changed:
- `path` — ...
Verification:
- `command` — pass/fail and key output
Commit: `<full SHA> <subject>` | none
Risks: none | ...

The coordinator—not the worker—updates todo or review state. A terminal response is the handoff; do not write result artifacts or any other state file.

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.

Reviews can't be read right now.

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.