agentleFS
nesquenanesquena/steward-os/skills/project-system-setup/SKILL.md

project-system-setup

Skill76 starsChanged 9 days agoMITView on GitHub

Run the setup interview to wire Steward to a specific project. Use when first installing the system, or when the project's repos/platforms/CI change. Produces config.yaml.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    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:

    Interview the human, section by section: Ask only what's needed; explain *why* each section matters (cite the playbook it feeds).

    line 19
  2. Distinctive:

    Confirm each answer: before writing it. Never guess — an unanswered question becomes a documented TODO in the output, not an assumption.

    line 24
  3. Distinctive:

    If the human pastes a token, refuse it and explain the secret-isolating helper pattern.

    line 38
  4. Distinctive:

    Don't over-automate at setup: Default new actions to Band B/C.

    line 41
  5. Very specific:

    Every block the human confirmed as not applicable is listed under not_applicable and is deleted or {}.

    line 49
  6. Distinctive:

    Every required section is filled or explicitly marked TODO.

    line 51

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

project-system-setup

When to load: first install of the system into a project, or when something structural changes (new repo, new chat platform, new CI, new review tool).

What it does: works through the setup interview conversationally and produces the project's config.yaml from the template.

Steps

  1. Read setup/setup-interview.md and setup/config.template.yaml. The interview is the question script; the template is the output shape.
  2. Interview the human, section by section. Ask only what's needed; explain why each section matters (cite the playbook it feeds). Accept "skip / not yet" for any optional section.
    • Required: project shape (Section 0), identity & repos, scope anchors, CI/tests, autonomy posture, secrets (if any public write).
    • Optional: chat/community, mentions, scheduled jobs.
  3. Confirm each answer before writing it. Never guess — an unanswered question becomes a documented TODO in the output, not an assumption.
  4. Write config.yaml from the template with the collected answers. Validate it parses, then run bin/config-lint on it from the steward-os checkout (the quickstart copies only skills/, so name the checkout path). The lint needs Ruby. A lint that can't run becomes a TODO, never a pass.
  5. Emit a TODO list of skipped sections and a recommended starting job set based on the autonomy posture (conservative default: scoreboard + digests at Level 2; nothing public-writing until a watchdog exists). If Section 0 left scoreboard-refresh disabled for want of a location, keep it disabled and carry the TODO.
  6. Point the human to adoption levels so they pick a starting level and know how to climb.

Pitfalls

  • Don't put secrets in config. Record the path to the locked credentials file, never a token. If the human pastes a token, refuse it and explain the secret-isolating helper pattern.
  • Don't over-automate at setup. Default new actions to Band B/C; autonomy is earned via the autonomy ladder, not granted at install.
  • Scope anchors are the highest-value answer — if the human is vague here, push for specifics, because the whole fit screen depends on them.

Verification

  • config.yaml exists and parses.
  • bin/config-lint exits 0 on it, or its failure to run is a TODO.
  • Every block the human confirmed as not applicable is listed under not_applicable and is deleted or {}.
  • Every required section is filled or explicitly marked TODO.
  • The human knows their starting adoption level and the next rung.

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.