agentleFS
MrHashMeMrHashMe/efficiency-playbook/skills/setup/SKILL.md

setup

Skill0 starsChanged 44 days agoMITView on GitHub

Install, upgrade or remove the Claude Code efficiency setup at user, project or local scope. Shows every change with its token cost and evidence, and asks before making it. Run only when the user types /efficiency-playbook:setup.

  • Installs packages

191 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use add-voice-transcription, add-discord and 10 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.
  • juancruzmunozalbeloSkill

    add-voice-transcription

    Add voice message transcription to NanoClaw using OpenAI's Whisper API. Automatically transcribes WhatsApp voice notes so the agent can read and respond to them.

    Also in 5 projects that use this

  • MicrosoftSkill

    add-discord

    Add Discord bot channel integration via Chat SDK.

    Also in 11 projects that use this

  • MicrosoftSkill

    add-slack

    Add Slack channel integration via Chat SDK.

    Also in 11 projects that use this

  • juancruzmunozalbeloSkill

    add-telegram-swarm

    Add Agent Swarm (Teams) support to Telegram. Each subagent gets its own bot identity in the group. Requires Telegram channel to be set up first (use /add-telegram). Triggers on "agent swarm", "agent teams telegram", "telegram swarm", "bot pool".

    Also in 7 projects that use this

  • +8 more from microsoft

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 171 edited versions we read keep these.

  • 1

    Distinctive

    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:

    Run inspect and say in two lines what it found: the Claude Code version, existing installs per scope, and any v6 or v7 setup that the plan …

    line 18
  2. Very specific:

    Ask which groups to apply (multi-select, default rows preselected) and which options to add.

    line 23
  3. Distinctive:

    verify problems, with their fix hints.

    line 31
  4. Very specific:

    Run uninstall --scope <scope> and show every row: what is removed, and what goes back to its value from before the install.

    line 36
  5. Very specific:

    After a yes, apply that plan as in step 4 above, then run verify --scope <scope>.

    line 37

“Common” is measured: kept by at least 6 in 10 of the 171 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

Arguments: $ARGUMENTS

The engine does every read and write; this skill only talks with the user. Run it as node "${CLAUDE_PLUGIN_ROOT}/engine/ep.mjs" <command> --json and read the envelope (ok, data, findings, error). Exit codes: 0 done, 1 findings, 2 usage, 3 blocked, 4 stale plan. Never edit the files it manages yourself, never write the plugin's own path into a file, and never commit or push.

First

  1. Run node --version. If Node is missing or older than 22, stop and walk the user through an install without admin rights, then wait until node --version reports 22 or later. Windows: download the x64 .zip of the current LTS from https://nodejs.org/en/download, compare Get-FileHash with the SHASUMS256.txt beside it, unpack it under %LOCALAPPDATA%\Programs, and add the folder that holds node.exe to the user PATH (Start, then "Edit environment variables for your account"), then open a new terminal. winget's Node package installs for all users and asks for admin, so don't offer it. macOS: brew install node if Homebrew is there. Otherwise, and on Linux: the official .tar.xz, checked against SHASUMS256.txt, unpacked under ~/.local, with its bin folder on PATH. Never pipe a downloaded script into a shell.
  2. Take the scope from the arguments, or ask, with three options:
    • user: every repository on this machine, for you only (files in your Claude config folder).
    • project: this repository, for everyone who clones it (files you review and commit).
    • local: this repository, for you only (git-ignored files). Project and local need a git repository. If this folder isn't one, say so and offer user scope.
  3. Run inspect and say in two lines what it found: the Claude Code version, existing installs per scope, and any v6 or v7 setup that the plan will migrate.

Install or upgrade

  1. Run plan --scope <scope>. If a finding is EP-ANSWER-NEEDED (project or local scope), find the repository's quick check and full acceptance commands in its package.json scripts, Makefile, justfile, pyproject.toml, Cargo.toml or CI workflow. Let the user confirm or correct them. Write {"commandMap":{"quick":["..."],"full":["..."]}} to a file in your scratchpad or the system temp folder, never in the repository, and run plan again with --answers <that file>.
  2. Show the rows grouped by their group value, each with a plain label: core is the rules file, migration removes a v6 or v7 setup, and every other value is a component id, such as lsp (code navigation), concise (output style), or context7 (docs lookup). For each row give the change in plain words, the target, alwaysLoadedTokens, the verdict with its evidence ids where it has one, and the egress. Whenever the egress isn't none, say what is sent and to whom. Show conflict and blocked rows apart, with before and after; they're never approved by default. Then list data.options, the components that are off by default, with the same facts.
  3. Ask which groups to apply (multi-select, default rows preselected) and which options to add. If the answer changes the plan, run plan again with --with <ids> or --without <ids>, and show what changed.
  4. Whatever the permission mode, wait for the user's answer to step 3 before this step; never choose for them. Run apply --scope <scope> --plan <hash> --approve <selector>, where the selector is default, group:<value> entries with those same values, or row ids, comma-separated. On exit 4, show the fresh plan and ask again. On exit 3, report error (its code and message) and stop.
  5. For each EP-MANUAL-STEP finding, give the user the exact command to paste, one per line.
  6. Run verify --scope <scope>.
  7. Report in at most 15 lines:
    • what changed, per group;
    • the always-loaded tokens added at this scope;
    • anything that now sends data off the machine;
    • verify problems, with their fix hints;
    • the undo command, written out in full the way you ran the engine: node "<plugin root>/engine/ep.mjs" rollback --scope <scope> --last, with the real path. That path changes when the plugin updates; /efficiency-playbook:setup uninstall always works. For project scope, add the git status lines and the suggested git add list from the apply result; the user commits. End with: "After each update of this plugin, run /efficiency-playbook:setup again."

Argument uninstall

  1. Run uninstall --scope <scope> and show every row: what is removed, and what goes back to its value from before the install. Files the user edited after the install are conflicts. Leave them unless the user picks them by id.
  2. After a yes, apply that plan as in step 4 above, then run verify --scope <scope>.
  3. Give the paste lines from the EP-MANUAL-STEP findings for plugins this setup installed. To remove this plugin as well, the last line is claude plugin marketplace remove efficiency-playbook, which also uninstalls it. In Desktop, + → Plugins → Manage plugins removes the plugin but keeps its marketplace. Then fully quit and reopen Desktop.

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.