agentleFS
theishandubeytheishandubey/dotclaude/skills/plan-execution/SKILL.md

plan-execution

Skill2 starsChanged 27 days agoMITView on GitHub

How workers execute a handoff plan or spec from the orchestrator - drift check, verify gates, scope, STOP conditions, approvals and the report. Preloaded by implementer, parallel-implementer, fixer and doc-writer.

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. 0 are common, 11 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 11

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

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

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

    When your task is a plan from plans/ (or any spec the orchestrator sends):

    line 9
  2. Distinctive:

    Run its drift check first.

    line 11
  3. Distinctive:

    If an in-scope file no longer matches Current state, stop and report.

    line 12
  4. Distinctive:

    Run every step's Verify and confirm the expected result before the next step.

    line 13
  5. Very specific:

    Stop before any step marked REQUIRES HUMAN APPROVAL and report it instead of running it.

    line 16
  6. Distinctive:

    Do everything the plan asks, across every site its pattern matches, and nothing it does not.

    line 17
  7. Distinctive:

    Read the files you will touch and one similar existing example first.

    line 22
  8. Distinctive:

    Before reporting done, run a real check that exercises the change: the project's type-checker, build or linter, or the changed command …

    line 23
  9. Distinctive:

    If only the project's declared dependencies are missing, install them with its own package manager and lockfile, never with sudo or a …

    line 25
  10. Distinctive:

    If no real check can run, say which one you did not run and why instead of reporting done.

    line 26
  11. Distinctive:

    Never report done with a red build, except for failures your change did not cause.

    line 27
  12. Distinctive:

    When working from review findings, address each exactly as specified.

    line 32
  13. Distinctive:

    If you disagree with one, implement it anyway when harmless, otherwise report the disagreement with your reasoning.

    line 33

“Common” is measured: kept by at least 6 in 10 of the 5 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

Executing a plan

When your task is a plan from plans/ (or any spec the orchestrator sends):

  1. Run its drift check first. If an in-scope file no longer matches Current state, stop and report.
  2. Run every step's Verify and confirm the expected result before the next step.
  3. Scope and STOP conditions are binding. On a STOP condition, stop and report; do not improvise around it.
  4. Stop before any step marked REQUIRES HUMAN APPROVAL and report it instead of running it.
  5. Do everything the plan asks, across every site its pattern matches, and nothing it does not. Say which sites you covered. Suggest extra features, files, docs or refactors in your report instead of doing them.
  6. On genuine ambiguity, or a plan that seems wrong, stop and ask in your report rather than picking a reading.
  7. The orchestrator maintains plans/README.md: never edit it, and ignore an untracked plans/ or advisor-plans/ in git status.
  8. Read the files you will touch and one similar existing example first; match its naming, error handling, imports and layout, but not its comment density.
  9. Before reporting done, run a real check that exercises the change: the project's type-checker, build or linter, or the changed command itself, plus the tests your agent instructions call for. Run no other tests. A syntax-only check, or a command that failed to start, does not count. If only the project's declared dependencies are missing, install them with its own package manager and lockfile, never with sudo or a system package manager. If no real check can run, say which one you did not run and why instead of reporting done.
  10. Never report done with a red build, except for failures your change did not cause. Fix an unrelated lint, type or build error only when it is in a file you already change and the fix is small, and list it under deviations. Report every other one with the exact output.
  11. Report against the plan's Done criteria.

When working from review findings, address each exactly as specified. If you disagree with one, implement it anyway when harmless, otherwise report the disagreement with your reasoning; never skip one silently.

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.