agentleFS
VforValVforVal/lumberjack/skills/setup/SKILL.md

setup

Skill0 starsChanged 24 days agoMITView on GitHub

Run-once interview that writes lumberjack's config, git hook and harness pinning for this repository

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

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

  • 0

    Common

    Most of the 171 edited versions we read keep these.

  • 2

    Distinctive

    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. Very specific:

    Check python3 --version, then python --version: use the first name that reports 3.11 or newer in place of python below.

    line 13
  2. Distinctive:

    If git is missing, stop and point to https://git-scm.com/downloads before running the helper.

    line 15
  3. Distinctive:

    Write the answers to a temporary file and run it once:

    line 99

“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

lumberjack setup

Interview the maintainer, then write the config once. The helper does all writing; never commit.

1. Discover

Check python3 --version, then python --version: use the first name that reports 3.11 or newer in place of python below. If neither does, stop and point to https://www.python.org/downloads/ . If git is missing, stop and point to https://git-scm.com/downloads before running the helper.

Resolve the helper at ../../scripts/setup_lumberjack.py relative to this skill's directory; use its absolute path as HELPER, quoted for the shell. Run it from any checkout and keep the JSON. Apply writes to repo, the primary checkout; when started_in is set, say so once.

python "HELPER" discover

2. Interview

One section at a time. Open each with the recommended answer taken from the JSON and a one-line reason; the maintainer accepts with one word or corrects it. Do not show the next section until this one is settled. If config.source is set (a re-run), recommend the values already in config instead of the guesses below.

0. Scope (scope). Ask first: local (just me: config in .git/lumberjack.toml, block in CLAUDE.local.md, nothing to commit, hook plain or none) or team (committed files). Recommend local when owners_guess has more than one name or on a first try (not when config.source is a committed file, which would win), team once agreed.

A. Harnesses (harnesses). Recommend claude when harness.claude_md or harness.claude_settings is true, codex when harness.codex_dir is true, both when unsure. With claude and team, also ask pin_claude_settings (default true): it writes the plugin into .claude/settings.json so teammates are prompted to install it. local: always false. instructions (team): instructions.target; if instructions.symlink, name links_to, never write it, and offer another of CLAUDE.md, .claude/CLAUDE.md, AGENTS.md that is no link.

B. Branch shape.

  • ticket_prefix: tracker.prefix, else ticket_prefix_guess, else "".
  • branch_shapes: propose the branch_shapes_guess shapes that cover most branches, with counts, as the canonical list; the maintainer adds, drops or edits. [] means <owner>/<ticket>-<slug>.
  • owners (only when a shape holds <owner>): owners_guess, else [] (any lowercase owner).
  • prototype_prefix: "prototype/". primary_branch: origin.default_branch, else "main".

C. Worktree roots (worktree_roots): only where to look for leftover folders; registered worktrees are swept wherever they live, so confirm, do not explore. Recommend worktree_parents as globs; outside the repo, a root without {name} only matches <dir>/<name>. Always write a list: the defaults scan the home directory.

D. Behaviour. cleanup: recommend "prune" (removes landed worktrees at session start); offer "report" to watch first. stop_gate: recommend true (blocks branch and primary debt).

E. Git hook (hook). Recommend hook_manager when set (husky, lefthook, pre-commit, simple-git-hooks), else plain; none skips the hook. The hook runs cleanup after every merge. An existing post-merge hook is kept: a marked block is appended to it, or it is chained when it is not a shell script. lefthook always gets a lefthook-local.yml job, even when lefthook.yml has its own post-merge section.

F. Evidence directory (evidence_dir). Ask only when config.evidence_dir is set, or a directory named .evidence, evidence or .<name> sits at the root and looks like scratch or handoff output. Its contents block cleanup of a worktree. Otherwise set "" without asking.

G. Doc roots (doc_roots). Always ask. Recommend config.doc_roots on a re-run, else the doc_root_candidates paths with their reasons, else ask where work documents (handoffs, plans, research, prototypes) live ("none" is []). Files there keep a landed worktree, reason docs.

H. Durable roots (doc_reference). Ask per doc root: durable (bundles are reference, never trashed by rule, only promoted or kept) or working (bundles may be trashed once their tickets land, under a saved doc_strategy). Never assume: one folder name is durable here, working elsewhere. Recommend durable when config.doc_reference maps the root to ["*"] or its name holds research, prototype or archive. Start from config.doc_reference: set a durable root to ["*"]; for a working one remove "*", deleting the key when no word is left. Copy config.doc_strategy as is.

3. Confirm

Show one consolidated summary and wait for an explicit yes:

  1. The full answers JSON. Keys: scope, harnesses, pin_claude_settings, instructions, branch_shapes, ticket_prefix, owners, prototype_prefix, primary_branch, worktree_roots, cleanup, stop_gate, hook, evidence_dir, doc_roots, doc_reference, doc_strategy.
  2. The files apply will write: team: .lumberjack.toml, the hook file, .claude/settings.json when pinning, the instructions block, .gitignore entries for roots inside the repository; local: .git/lumberjack.toml, .git/hooks/post-merge, CLAUDE.local.md, .git/info/exclude. Apply prints wrote <file> for each one it touched.
  3. The git config commands: git config core.hooksPath .githooks (team plain hook, only when hooks_path is unset) and git config lumberjack.script <plugin's sweep_worktrees.py>.

Any change requested means back to the affected section, then confirm again.

4. Apply

Write the answers to a temporary file and run it once:

python "HELPER" apply ANSWERS.json

Relay its output lines verbatim. On a setup: error, fix the named answer and run again. Team: tell the maintainer to review git diff and commit it themselves.

5. After apply

  • Relay printed, not run: lines for each selected harness. With Codex, say that trusting the hooks (/hooks in Codex) is a one-time step per machine.
  • gh missing or unauthenticated...: tell them cleanup still works but decides "landed" from git alone; run gh auth login for pull-request detection.
  • note: run ... to activate the hook: relay it; the maintainer runs that command.

Re-running

Safe at any time. discover returns config with source set (.git/lumberjack.toml means local); recommend those values. Apply rewrites only the config, the values it owns in .claude/settings.json and text between its marker lines, and only appends ignore lines.

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.