agentleFS
cochinescucochinescu/decent-skills/skills/init-repo-docs/SKILL.md

init-repo-docs

Skill8 starsChanged 56 days agoMITView on GitHub

Backfill STATUS.md + TODO.md (and DECISIONS.md if warranted) for an existing repo from git history, plans, and the wiki.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run from the repo root. Goal: bootstrap STATUS.md, TODO.md, and optionally DECISIONS.md for a repo that predates that convention.

    line 11
  2. Distinctive:

    If README or CLAUDE.md points at a project wiki or history page, read it.

    line 24
  3. Distinctive:

    READ-ONLY — never edit wiki content from here.

    line 26

≈ “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

User-invoked. Run only when the user asked for this skill. Target: $ARGUMENTS (optional scope hint, e.g. "last 6 months" or "only STATUS"; empty = full backfill of this repo)

Run from the repo root. Goal: bootstrap STATUS.md, TODO.md, and optionally DECISIONS.md for a repo that predates that convention. Read-only toward everything except those three files; never commit.

Step 1 — Mine sources (delegate bulk reading if a cheaper worker is available; the main session adjudicates and writes):

  • git log: dates, merges, release tags, bursts of activity → milestone candidates. Use commit dates as the source of truth for entry dates.
  • Repo CLAUDE.md + README: current capabilities, stated conventions.
  • Planning docs (planning-docs/ or the repo's detected plan dir): dated plans → compare plan vs what actually landed; unimplemented plan items are TODO candidates.
  • If README or CLAUDE.md points at a project wiki or history page, read it. If a local wiki page exists for this repo, read that too. READ-ONLY — never edit wiki content from here.
  • In-code markers: TODO/FIXME/HACK comments → TODO candidates (with file:line).
  • gh: open PRs/issues and non-default branches → TODO candidates.

Step 2 — Categorize. Every entry must trace to evidence (a commit, file, plan doc, or wiki line); never invent history or rationale.

  • STATUS.md: things that HAPPENED — shipped features, milestones, migrations, fixes worth remembering. Dated yyyy-mm-dd, newest first, one short block each (what changed, where, outcome). For a long history, summarize older periods into coarse per-month/quarter blocks instead of per-commit noise; recent work gets finer grain.
  • TODO.md: things still OPEN — unimplemented plan items, wiki open threads, code markers, open PRs/branches, known gaps. Prioritized checklist; each item notes its source. Drop items that evidence shows already landed (that's what the cross-check is for).
  • DECISIONS.md: ONLY if the mining surfaces real architectural choices with a recoverable "why" (from plan docs, wiki, commit messages). One line each: date, decision, why. If the why is not recoverable, write "(rationale not recorded)" — never fabricate. Fewer than ~3 solid entries → skip the file entirely.

Step 3 — Write. If any of the files already exist, merge into them (preserve existing entries, keep date ordering) — never clobber. Then report a compact summary: entry counts per file, the top open TODOs, and anything ambiguous that needs my judgment (e.g. "branch X looks abandoned — TODO or drop?"). Stop there; do not commit.

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.