agentleFS
doivamongdoivamong/agent-workbench/.claude/skills/awb-uninstall/SKILL.md

awb-uninstall

Skill2 starsChanged 6 months agoMITView on GitHub

WHAT: drive the kit's own uninstaller in plain language — show what would be removed first (dry-run), get an explicit yes, then remove, and say honestly what is KEPT (the files you edited) rather than claiming a clean wipe. USE WHEN: a user wants to take the workbench back out of a project ("remove agent-workbench", "uninstall the kit", "take the hooks out", "undo the install"). DO NOT TRIGGER: installing or verifying guards (that's awb-install-and-verify); hand-deleting files yourself (this skill only ever calls uninstall.py); a general "is my code good?" review (that's awb-review).

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Does: run uninstall.py <project> (dry-run) to get the removal plan, translate it into plain language, get an explicit confirm, then run …

    line 28
  2. Very specific:

    Dry-run first — show, don't touch: Run python uninstall.py <project-path>.

    line 45
  3. Very specific:

    HARD GATE: explicit confirm before applying: Ask the user to confirm they want to proceed.

    line 52
  4. Very specific:

    No manifest → stop, don't guess: If uninstall.py reports no installer-manifest, it refuses to pattern-delete blindly — relay that and the …

    line 57
  5. Distinctive:

    Only an untouched install → uninstall leaves git status fully clean.

    line 67

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Uninstall the workbench, honestly

Announce on activation: "Using awb-uninstall — I'll show you exactly what would be removed first, then remove it only once you confirm."

The persona this kit serves often can't read a manifest or diff a settings file, so they can't tell a clean uninstall from one that quietly deleted work they'd changed. The only honest way to remove the kit is to show the plan before touching anything, get a clear yes, and then state plainly what was removed and what was kept. uninstall.py is built for exactly this — it is dry-run by default and never deletes a file you edited. This skill makes the agent use it that way instead of improvising file deletions.

Scope

  • Does: run uninstall.py <project> (dry-run) to get the removal plan, translate it into plain language, get an explicit confirm, then run uninstall.py <project> --yes and relay what was removed vs. kept.
  • Does NOT: hand-delete any file (only uninstall.py removes anything); remove files you edited since install (those are KEPT by design — see Honesty); guess what to remove when there's no installer-manifest (uninstall.py refuses, and so should you). It is a bypassable exemplar, not a gate.

Precondition

uninstall.py lives in the kit folder, not in the project it was installed into. Run it from the kit checkout, pointing at the adopter project: python uninstall.py <project-path>. (Same precondition the install skill documents — the uninstaller is never copied into adopter projects.)

Process

  1. Dry-run first — show, don't touch. Run python uninstall.py <project-path>. With no --yes it changes nothing; it prints the plan: which files would be removed, which are KEPT (modified since install), the settings.json revert, and the .gitignore cleanup.
  2. Relay the plan in plain language. Tell the user, in their terms: how many files would be removed, which (if any) are kept because they edited them, and that the hooks will be unwired from settings.json. Don't paraphrase a plan you didn't read — read the actual output.
  3. HARD GATE: explicit confirm before applying. Ask the user to confirm they want to proceed. Do not run --yes on assumption. If they hesitate or say "what gets kept?", answer from the dry-run output before going further.
  4. Apply. Once confirmed, run python uninstall.py <project-path> --yes. Relay the result honestly: N files removed, M kept (modified), empty dirs pruned.
  5. No manifest → stop, don't guess. If uninstall.py reports no installer-manifest, it refuses to pattern-delete blindly — relay that and the remedy it prints (re-run install to regenerate the manifest, or remove by hand). Never substitute your own rm.

Honesty (say this, don't soften it)

  • "Remove" does not mean "wipe everything." A file you edited since install (its bytes no longer match the kit's recorded sha256) is KEPT, not deleted — uninstall never destroys your changes. Say which files were kept and that they're left for the user to remove by hand if they truly want them gone.
  • Only an untouched install → uninstall leaves git status fully clean. If the user edited copied files, the tree won't be empty afterward — and that's correct, not a bug.
  • The uninstaller reverts only the hook commands install added; hooks the user added themselves survive. Don't claim a blanket "settings.json restored to factory."

Anti-rationalization

You'll thinkReality
"They said uninstall, I'll just run --yes"Dry-run first and show the plan. --yes without a confirmed plan can surprise them with what was (or wasn't) removed.
"I'll just rm the kit files myself, it's faster"Then you bypass the sha-check that keeps their edited files — exactly the safety this skill exists for. Only uninstall.py removes anything.
"No manifest, I'll figure out what to delete"That's the blind pattern-delete uninstall.py refuses on purpose. Relay the refusal; never guess.
"I'll say it removed everything"If any file was edited it was KEPT. Saying "removed everything" is the dishonest over-promise this skill forbids.

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.