agentleFS
cbdreamer11cbdreamer11/CB-loop-kit-claude-plugin/skills/loop-close/SKILL.md

loop-close

Skill8 starsChanged 4 months agoMITView on GitHub

Closes a slice or a session - adversarial audit of the diff, ledger entry, selective commit, honest report, and the handoff for the next session. Use before considering anything finished.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    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:

    If the auditor finds a real bug, also look for the class of that bug: search for its siblings elsewhere in the codebase and close them in …

    line 17
  2. Distinctive:

    If something failed, say it failed and show the output.

    line 50
  3. Very specific:

    Leave .loop/STATE.md so the next session needs no instructions: the next unchecked item is next, and the command to run it is written …

    line 55

≈ “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

Close

Closing is not paperwork. It is the step that makes the next session cheap and stops the same work from being done twice.

1 · Audit before you close

Dispatch the loop-auditor agent on the full diff, with the item's non-negotiables from .loop/STATE.md. Its job is to refute. Iterate until there are zero CONFIRMED findings — and do that before anything is published, not after.

If the auditor finds a real bug, also look for the class of that bug: search for its siblings elsewhere in the codebase and close them in the same session, or record them as their own items. One fixed instance and three live ones is not a fix.

2 · Commit

  • Stage only your own files. Never git add -A when someone else has uncommitted work in the tree.
  • A descriptive message: what changed and why, not "fixes".
  • On the working branch. Never the protected branch — publishing is the owner's action, and it stays that way even when it is obviously fine.
  • Team mode: open the pull request, make sure CI is green, assign a reviewer. Done means merged, not "it works on my machine".

3 · Write the record

  • .loop/STATE.md — item to [x] plus one line of what was verified and how. Anything unverified becomes GAP: <why>. Anything blocked becomes PARKED: <why> + exactly what is missing.
  • .loop/DECISIONS.md — any real choice made, with the rejected alternative.
  • .loop/GOTCHAS.md — anything that fooled you: the symptom, why it was convincing, and how to detect it next time. This file is the compound interest of the method.

4 · Report honestly

Three lists, no softening:

  • Verified — with what you observed for each, in one line.
  • Gaps — what is not verified and why (a missing access, no browser, no test environment). Never dress a gap as done.
  • Blocked on the owner — a key, real money, a business or legal decision, or permission to publish. State the exact thing needed, not a vague ask.

If something failed, say it failed and show the output. A report that reads better than the work is the one failure this method cannot recover from.

5 · Hand off

Leave .loop/STATE.md so the next session needs no instructions: the next unchecked item is next, and the command to run it is written beside it. Print that command.

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.