agentleFS
TarasTsavolykTarasTsavolyk/claude-code-frontend/.claude/skills/prune/SKILL.md

prune

Skill1 starsChanged 4 months agoMITView on GitHub

Remove kit capabilities this project won't use — agents, skills, and rules — and fix every cross-reference left behind. Manual, destructive, graph-aware. Run deliberately; it is never auto-triggered and commits on a branch so git is the undo.

14 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use deliver, wait-what and 4 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • GoogleSkill

    deliver

    Posts a condensed version of the briefing to a Google Chat or Slack incoming webhook, so the daily run delivers itself — skips silently when no webhook is configured.

    Also in 3 projects that use this

  • AWSSkill

    wait-what

    Stop. That last message did not land: re-pitch it.

    Also in 3 projects that use this

  • NVIDIASkill

    research

    Before designing a feature, research how comparable libraries solve the problem, check best practices, known pitfalls, and upstream changes. Use when starting a new feature, investigating an unfamiliar area, or the user asks to research something.

    Also in 5 projects that use this

  • GoogleSkill

    prototype

    Build a throwaway prototype to answer a design question. Use when the user wants to sanity-check whether a state model or logic feels right, or explore what a UI should look like.

    Also in 4 projects that use this

  • BlockSkill

    plan

    Draft a TDD, PRD, or lightweight plan in the conversation without committing it.

    Also in 3 projects that use this

  • AWSSkill

    ship

    Finalize dev work on an issue branch — update docs, push, and open a PR.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 0

    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:

    Guard the working tree: Run git status --short. If it is not clean, stop and ask the user to commit/stash first, or to explicitly approve …

    line 21
  2. Very specific:

    Re-detect & summarize: If CLAUDE.md changed, re-run node .claude/scripts/detect-stack.mjs.

    line 44

“Common” is measured: kept by at least 6 in 10 of the 13 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

Prune the kit

Removes whole capabilities (agent + skill + rule bundles) the project doesn't need, then reconciles every file that referenced them. Destructive and deliberate — there is no auto-trigger; the user invokes it by name.

Two hard rules

  • Never touch CHANGELOG.md. It is append-only release history; past entries legitimately name things that no longer exist.
  • Never run on a dirty tree. git is the primary undo (step 1); /rewind plus file checkpointing is a second net, but it is session-scoped — don't lean on it in place of a clean tree.

Procedure

  1. Guard the working tree. Run git status --short. If it is not clean, stop and ask the user to commit/stash first, or to explicitly approve proceeding anyway. Recommend a dedicated branch (e.g. chore/prune-kit).
  2. Ask what to remove — use AskUserQuestion checkboxes, not a "type 1,2,3" list. Present the removable capabilities as multiSelect: true questions so the user ticks boxes. The tool caps each question at 4 options and each call at 4 questions, and there are 12 units — so issue one call with these four checkbox questions (3 + 4 + 2 + 3):

    • Tier A — workflows: e2e tests · from-figma · release
    • Tier A — domain rules: i18n · forms · data fetching · config
    • Tier A — platform rules: error handling & observability · ssr (ssr is a default-remove unless metaFramework in facts says the project server-renders)
    • Tier B — high blast radius ⚠️: security · performance · accessibility

    Put each unit's one-line effect in its option description; for every Tier B option, lead the description with its ⚠️ warning. After the user submits, for any Tier B pick re-confirm explicitly before deleting (they're woven through many files and two of them back a stated stance in CLAUDE.md). Never offer the Tier C spine.

  3. Delete each chosen unit's files (unit map below). Special case — e2e tests is a trim, not a delete: remove the "E2E (Playwright)" section from rules/testing.md and the test:e2e bullet from .claude/skills/verify/ step 1; Keep test-engineer/testing.md.
  4. Reconcile the universal hubs — touched by almost every removal:
    • .claude/rules/workflow.md — remove the agent from any flow / Quality-Gate line; if a whole flow section empties out, delete that section.
    • CLAUDE.md — the removed unit's row in the The rules table, plus any structure/stack line that names it. If you remove security or accessibility, also drop the matching stance sentence under that table.
    • README.md / docs/ — kit repo only. A deployed project has only .claude/ + CLAUDE.md (the kit's README/docs/CHANGELOG aren't copied in), so skip these there. When pruning the kit repo itself: remove the name(s) from the Contents tree and drop any pipeline/design-note prose naming a removed agent. Recipe rows that delete docs/… or .github/… are likewise no-ops when those files aren't present.
  5. Reconcile by graph, not by memory. For every removed unit, pass its slug(s) as they appear in file names (data-fetching, security-scanner, from-figma):
    node .claude/scripts/check-refs.mjs <removed-name> [<removed-name>...]
    Review every hit (it prints file:line [name] text). Fix real dangling references. A hit that is a legitimate English word (e.g. "release") or a generic concept mention may be left — but only after confirming it isn't actually a reference to the removed unit. Re-run until clean or every remaining hit is justified. Typical danglers: (see <rule>.md) clauses in sibling rules and agent bodies, the quality-gate board lines in workflow.md, and the unit's row in CLAUDE.md's The rules table.
  6. Re-detect & summarize. If CLAUDE.md changed, re-run node .claude/scripts/detect-stack.mjs. Tell the user exactly what was deleted and which files were edited; show git diff --stat.
  7. Commit. On a branch, one focused commit (never main — see CLAUDE.md → Git). Surface the changed-file list and the full commit message for approval before committing.

Unit map (what each unit deletes)

Reference cleanup is discovered by step 5's graph pass, not memorized here — this table only says which files go.

UnitFiles
e2e tests(trim, don't delete — see step 3)
from-figmaskills/from-figma/
releaseskills/release/ · docs/release-automation.md · .github/workflows/release.yml
error handling & observabilityrules/error-handling.md
i18n · forms · data fetching · config · ssrthe matching rules/<name>.md
security ⚠️rules/security.md · agents/security-scanner.md
performance ⚠️rules/performance.md · agents/performance-auditor.md
accessibility ⚠️rules/accessibility.md · agents/accessibility-auditor.md

Tier B ⚠️ re-confirmation: security and accessibility each back a stance CLAUDE.md states outright — the server-trust-boundary rule and "a requirement, not a nice-to-have". Removing the rule means retracting the stance; spell that out when re-confirming.

Tier C — never removable here: the quality-gate board (ui-reviewer, test-engineer), the onboarding/verification skills (wizard, prune, verify, scaffold-component), and the spine rules (code-style, architecture, styling, testing, plus the global workflow). Removing these would gut the kit.

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.