agentleFS
mantaray0mantaray0/crew/skills/crew-state/SKILL.md

crew-state

Skill1 starsChanged 5 months agoMITView on GitHub

Format of crew's `.planning/` directory — ROADMAP markers, milestone plans, backlog items, LOG lines, config. Use when reading or writing any `.planning/` file, inside or outside a `/crew:*` command.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 1 is distinctive, and 1 is 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.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When an idea moves into the roadmap or is dropped, delete its file.

    line 83
  2. Very specific:

    Put each decision to the user with a recommended answer they can accept as is.

    line 132

≈ “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

crew state

crew keeps a project's state in .planning/, committed with the project. A fresh session orients from these files alone, so every change a command makes lands here before the command ends.

FileHolds
PROJECT.mdWhat the project is, its decisions and their why, current state, stack, ## Checks, ## Reference
ROADMAP.mdMilestones → phases with status markers
plans/<n>_<slug>/One folder per milestone: _spec.md (intent) + one plan file per phase
backlog/One Markdown file per idea
LOG.mdAppend-only history
config.jsonThe few settings below
archive/Closed milestones
reference/Long-form docs, read only when PROJECT.md's ## Reference points there for the task at hand

Projects grow their own files (HANDOFF.md, research/, a single backlog/ideen.md, flat plan files). Read them when relevant and leave them as they are. Older projects also carry claims.json; it is no longer read or written.

Tolerant reading. Projects written by older crew versions differ in headings, id depth and folder names. Read what is there; write new content in the style the file already uses.

ROADMAP.md

Markers: [ ] open · [>] in progress · [x] done · [~] deferred.

## Meilenstein 3: Suche
- [x] 3.1 Index und Anschluss (erledigt 2026-10-02)
- [>] 3.2 Suchfeld (depends: 3.1)
- [ ] 3.3 Rückfrage (depends: 3.1, 3.2)
  • A milestone is any ## heading; Meilenstein N: and Milestone N — are both in use. A phase is a - [?] line under it; its id is the leading token (3.2, 29.9.15, or none).
  • Done stamp: append (erledigt YYYY-MM-DD), worded in language.files and matching earlier stamps in the file.
  • depends: names phases that must be [x] first. The next phase is the first [>], else the first [ ] whose dependencies are done.
  • Phases are edited as text: insert, reorder, defer, drop. Ids are labels, never renumbered.

plans/

plans/3_suche/
  _spec.md          ← milestone intent, written by /crew:brief, permanent
  3.1_index.md      ← phase plan, written by /crew:plan
  3.2_suchfeld.md
  • _ prefix = meta file (_spec.md, _roadmap.md), never a phase. Numbers join names with _; words inside a name with -; slugs are lowercase ASCII even when content is German.
  • Locate a phase plan by globbing plans/**/*.md for its id prefix.
  • _spec.md: Goal/problem · Requirements · Acceptance criteria · Out of scope · optional Notes for planning (structural ideas the brief parked for /crew:plan; input, never requirements).
  • Phase plan: the how for one phase, pointing at _spec.md for intent instead of copying it:
# 3.2 Suchfeld

> **Scope:** <deliverable and this phase's acceptance, 1–3 lines>. Intent: see `_spec.md`.

## Plan
- Files: …
- Tasks: action · pattern to mirror · how to check it
- Risks: the invariants a reviewer must not trade away

backlog/

One file per idea, backlog/<NNN>_<slug>.md; NNN is the highest existing number + 1, never reused. Read every backlog/*.md, whatever its name.

---
id: 012
title: <short name>
priority: medium        # low · medium · high
created: 2026-10-22
---
**Why:** <the motivation — required>
**Area:** <optional>

<free-form context>

When an idea moves into the roadmap or is dropped, delete its file; the roadmap and plan carry it from there.

LOG.md

Append only; earlier entries stay as written, however long. One line per event:

- 2026-10-22 · 3.2 · a1b2c3d · checks green · review: 3 fixed, 2 rejected
- 2026-10-24 · Meilenstein 3 closed · archive/plans/3_suche/
- 2026-10-25 · 3.4 · stopped · review finding needs a scope decision

Fields: date · phase · commit (or stopped) · verify result. Words follow language.files. Reasons for rejected findings and deviations go in the commit body, where they stay next to the code.

PROJECT.md

Free-form, but these sections are read by commands:

  • ## Checks: the commands verify runs, one per line (test, typecheck, lint). Without it, use the package.json scripts of those names and say which ones ran.
  • ## Reference: one line per reference/<topic>.md: link, what it covers, when to read it.
  • The current-state section, refreshed when a milestone closes.

Closing a milestone

A milestone is closable when every phase is [x] or [~]. With the user's OK:

  1. Refresh the current-state section in PROJECT.md.
  2. Move plans/<n>_<slug>/ to archive/plans/<n>_<slug>/ and write the milestone's ROADMAP section into it as _roadmap.md.
  3. Replace that section in ROADMAP.md with one line: ## Meilenstein N: <title> — archiviert YYYY-MM-DD → archive/plans/<n>_<slug>/_roadmap.md.
  4. Append the LOG line.

Existing archives in other layouts stay where they are.

config.json

{
  "language": { "files": "en" },                    // language of everything crew writes into .planning/
  "workflow": { "execute": {
    "loop": "all",                                  // "all": run the milestone's phases in sequence · "one": one phase, then stop
    "review": "medium"                              // /code-review effort after each phase: off · low · medium · high · xhigh · max
  } }
}

Resolve each key as project (.planning/config.json) → global (~/.claude/crew/config.json) → the defaults above; a value of "inherit" means the next layer down. At session start crew removes every other key from both files, with a backup under ~/.claude/crew/backup/. Commit format, branches and the issue tracker come from the project's CLAUDE.md.

Working with the user

  • Put each decision to the user with a recommended answer they can accept as is. Closed choices (2–4 options) go through AskUserQuestion, independent ones batched; open questions go in text.
  • Show what you will write, get the OK, then write.
  • Structure stays English and stable (markers, frontmatter keys, file names); prose follows language.files.

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.