Bootstrap agents-concerto in the current project by interviewing the user for every configurable option — target repos (with per-repo PR host), task_source, base branch, commit convention, and fix-cycle cap — then writing a filled .agent-workspace/config.md. Detects sensible defaults and asks for confirmation rather than making the user edit YAML by hand.
Scanned 9/6/2026
Install to Claude Code
npx -y skills add moruno21/agents-concerto --skill setup --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Setup?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/moruno21-setup)More formats (shields.io, HTML) on the badges page.
---
name: setup
description: Bootstrap agents-concerto in the current project by interviewing the user for every configurable option — target repos (with per-repo PR host), task_source, base branch, commit convention, and fix-cycle cap — then writing a filled .agent-workspace/config.md. Detects sensible defaults and asks for confirmation rather than making the user edit YAML by hand.
disable-model-invocation: true
---
# /setup — interactively configure the orchestrator for this project
Prepare the current project to run `/agents-concerto:run`. This creates only
**local** project config; the orchestrator logic stays in the plugin. Your job is
to **interview the user** for each configurable option and write a complete
`./.agent-workspace/config.md` — the user should not have to hand-edit YAML.
## How to ask
- Ask **one decision at a time**, and always **offer a default** (or a value you
detected) so the user can accept with a single word. Prefer structured choices
over free text where the option is an enum.
- **Detect, then confirm** — don't ask cold when you can infer. Sniff the repo
for its remote, default branch, and test command, propose what you found, and
let the user correct it.
- Offer an **"accept all defaults"** fast path up front for users who just want a
single-repo, `task_source: none` setup.
- Keep the two axes explicit whenever they come up: **`task_source`** is where
tasks are *read from*; a repo's **`host`** is where PRs are *opened*. They are
independent.
## Steps
1. **Workspace.** Ensure `./.agent-workspace/` exists in the current project.
2. **Don't clobber.** If `./.agent-workspace/config.md` already exists, do **not**
overwrite it silently — show its current contents and ask whether to edit it,
reconfigure from scratch, or leave it. Only proceed to write once the user
chooses.
3. **Interview — target repos** (the *implement* axis). Repos are a list; one or
N is the same to the pipeline. For each repo, gather:
- **`path`**: an absolute path. **Validate it exists on disk** and is a git
repo; if not, flag it and re-ask.
- **`slug`**: a short id. Propose one from the directory name; confirm.
- **`test`** (optional): a bounded suite command. Sniff the repo (e.g.
`package.json` scripts, `pyproject.toml`, `Makefile`) and propose one; let
the user accept, change, or skip it.
- **`host`** (optional): where PRs open for this repo — `github`, `gitlab`, or
`none` (branch-only). **Auto-detect from the repo's git remote** and propose
that (github.com/gitlab.com → that host; no usable remote → `none`); the
user confirms or overrides. Record `host` explicitly only when it differs
from what auto-detection would give, or when the user wants it pinned.
Then ask **"add another repo?"** and loop until done.
4. **Interview — `task_source`** (the *read* axis). Ask where tasks are read
from: `none` (default — supplied to `/run` or read from
`.agent-workspace/feature-request.md`), `github`, `gitlab`, `linear`, or
`jira`. Remind the user this is independent of `host` — e.g. read from Linear,
open PRs on GitHub.
5. **Interview — conventions & limits.** For each, propose the default and let
the user accept or change:
- **`base_branch`**: the ref feature branches are cut from. **Detect the
repo's default branch** (e.g. `main`/`master`) and propose it. If repos
disagree, ask per repo or pick the common one and note it.
- **`commit_convention`**: `conventional` (default) or `default` (free-form).
Either way the structural-then-behavioral split is enforced.
- **`max_fix_cycles`**: the review→fix cap before escalating to
`ready-for-human` (default `3`).
- **`branch_convention`**: default `feat/<task>-<repo-slug>`. Only ask if the
user wants to customize it; otherwise keep the default silently.
6. **Write the config.** Write `./.agent-workspace/config.md` from the collected
answers, following the structure and comments of
`${CLAUDE_PLUGIN_ROOT}/.agent-workspace/config.md.example`. Then **show the
final file** and ask the user to confirm it reads correctly.
7. **Gitignore reminder.** Remind the user that local runtime files (`config.md`,
`feature-request.md`, `plan.md`, `runs/`) are project-local and should be
gitignored in their project if they don't want them committed.
8. **Next step.** Point them at:
`/agents-concerto:run <task description>` (or, with `task_source: none`, run
`/agents-concerto:shape <rough idea>` first to generate
`./.agent-workspace/feature-request.md` with testable acceptance criteria).
Do not run the pipeline from here — `/setup` only prepares config.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!