roborew/setup-skills

Setup skills (per-repo)

Scaffold the configuration that grill-me, to-issues, debug-fix, and zoom-out assume:

• Issue tracker — where issues live (GitHub default; local markdown optional)
• Triage labels — strings for agent-ready vs human-needed work
• Domain docs — single CONTEXT.md + docs/adr/ vs multi-context CONTEXT-MAP.md

This is a prompt-driven skill: explore the repo, present findings, confirm with the user, then persist via scribe (Task scribe with target_path + full content). Do not invent remote URLs; read git remote -v when relevant.

Process

1. Explore

• git remote -v — GitHub? GitLab? none?
• Root README.md, AGENTS.md, CLAUDE.md — is there already an ## Agent skills block?
• CONTEXT.md, CONTEXT-MAP.md, docs/adr/, docs/agents/

2. Present and confirm (one topic at a time)

A — Issue tracker. Default: GitHub if origin is github.com. Alternatives: GitLab, local markdown under .scratch/issues/ (describe convention in issue-tracker.md), or freeform (Jira/Linear) as prose in issue-tracker.md.

B — Triage labels. Canonical five (map to repo labels if different):

| Role | Default string | |------|------------------| | Maintainer must evaluate | needs-triage | | Waiting on reporter | needs-info | | AFK agent can pick up | ready-for-agent | | Needs human implementation | ready-for-human | | Won't fix | wontfix |

C — Domain layout. Single-context — one CONTEXT.md at repo root + docs/adr/. Multi-context — CONTEXT-MAP.md at root pointing to per-context CONTEXT.md and optional per-context docs/adr/.

3. Draft for user edit

Show drafts of:

1. docs/agents/issue-tracker.md
2. docs/agents/triage-labels.md
3. docs/agents/domain.md
4. A short ## Agent skills markdown block (for AGENTS.md or README.md)

4. Write via scribe

Invoke scribe three times (or one batched sequence per parent rules) with:

• target_path: docs/agents/issue-tracker.md
• target_path: docs/agents/triage-labels.md
• target_path: docs/agents/domain.md

Then either:

• target_path: AGENTS.md — create or append ## Agent skills (if creating, include a one-line title at top: # Agent notes), or
• Append the same block to README.md if the user prefers no new file.

Never create both AGENTS.md and duplicate the block in README.md.

5. Done

Tell the user setup is complete. They can edit docs/agents/*.md anytime. Re-run this skill when switching issue tracker or domain layout.

---

Seed: docs/agents/issue-tracker.md (GitHub)

# Issue tracker

Issues for this repository are tracked on **GitHub**.

- **CLI:** `gh issue create`, `gh issue view`, `gh issue list`
- **Remote:** (fill from `git remote get-url origin`)

Non-GitHub workflows belong in this file as plain-English steps for agents.

Seed: docs/agents/triage-labels.md

# Triage labels

| Role | Label on this repo |
|------|---------------------|
| Needs triage | needs-triage |
| Needs info | needs-info |
| Ready for agent | ready-for-agent |
| Ready for human | ready-for-human |
| Won't fix | wontfix |

When creating issues, use the **Label on this repo** column. If a label does not exist yet, create it or omit and note in the issue body.

Seed: docs/agents/domain.md

# Domain docs for agents

## Layout

- **Mode:** single-context | multi-context
- **Glossary:** `CONTEXT.md` at repo root (or see `CONTEXT-MAP.md` for paths)
- **ADRs:** `docs/adr/` (system-wide); bounded contexts may use `<context>/docs/adr/`

## Consumer rules

1. Before naming entities in plans or issues, read the active `CONTEXT.md`.
2. Before changing architecture, scan `docs/adr/` for decisions in that area.
3. `CONTEXT.md` is glossary-only — not implementation specs.

Seed: ## Agent skills block

## Agent skills

### Issue tracker

Configured for this repo. See `docs/agents/issue-tracker.md`.

### Triage labels

See `docs/agents/triage-labels.md`.

### Domain docs

See `docs/agents/domain.md`.