Skip to content
Back to skills

Full Setup

ASecurity

Complete Agent Engineering Kit adoption in an existing repository. Use when asked to set up the kit end to end, make a repository agent ready, or finish onboarding after bootstrap.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 29, 2026
ai-agentspythonshellbashnodegitapi

Works with

  • api

Security analysis

A100/100

Scanned September 29, 2026

npx -y skills add Rvey/agent-engineering-kit --skill full-setup --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Full Setup?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Full Setup
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/rvey-full-setup/badge)](https://www.skillsdirectory.com/skills/rvey-full-setup)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: full-setup
description: "Complete Agent Engineering Kit adoption in an existing repository. Use when asked to set up the kit end to end, make a repository agent ready, or finish onboarding after bootstrap."
---

# Complete repository setup

The deliverable is a working local verification loop and, on GitHub, CI and
required checks. A copied template or green policy check alone is incomplete.

## Workflow

1. **Inspect the target.** Read its existing `AGENTS.md`, repo instructions,
   manifests, lockfiles, CI, hooks, and Git remote. Record real lint, type,
   test, and build commands. Read-only inspection comes before installation.
2. **Install safely.** If needed, run the kit's `scripts/bootstrap.sh` against
   the repository, with `--dry-run` first when the target has existing hooks,
   workflows, or instruction files. Existing files are preserved; a rerun
   fills missing skill files. Do not use `--force` on authored instructions,
   workflow files, or custom skills without comparing them first. If a file
   conflicts, merge it.
3. **Generate instructions and gates.** Follow
   `.agents/skills/agents-md/SKILL.md` to write project-specific `AGENTS.md`
   and executable `.agents/verify.sh`. Start `.agents/verify.sh` from the
   closest `.agents/verify-templates/` starter (shell, node, python) and
   replace every `SETUP:` line with a command traced to real config; an
   unedited starter is not a setup. Ask the user only for facts the repo
   cannot answer. Never invent commands or retain template placeholders. If a
   gate class has no real command yet, report that gap explicitly.
4. **Wire CI.** The copied `.github/workflows/gates.yml` runs the full
   verification script. Add only the toolchain setup the repo actually needs,
   including a reproducible dependency install from its lockfile. Keep policy
   and gates separate, and run both in CI. On a non-GitHub host, use its
   equivalent pipeline instead.
5. **Install the local fast gate.** Run `.agents/install-hooks.sh`
   after `.agents/verify.sh` works, or `.agents/setup-check.sh . --fix` to
   scaffold the missing pieces first. If another hook manager or hook already
   exists, integrate `bash .agents/verify.sh --fast` there instead of
   replacing it.
6. **Configure review ownership.** For GitHub, copy
   `.github/CODEOWNERS.example` to `.github/CODEOWNERS` with real handles.
   Derive those from existing CODEOWNERS or repository teams when possible;
   ask if ownership is unknown. Remove irrelevant example paths.
7. **Verify locally.** Run the full verification script, the fast hook gate,
   and `.agents/setup-check.sh .`. Fix failures before
   declaring local setup ready. Review the generated instructions for
   commands that have no config source.
8. **Verify remote enforcement.** With the user's repository access, require
   `gates / verification gates` and `policy / loop evidence` in branch
   protection, plus pull requests and code owner review where configured.
   Read settings back through the host API or UI; a workflow file alone does
   not make a check required. If access is missing, report the exact setting,
   repository, and reason it remains open. Never claim remote readiness from
   local checks alone.

## Handoff

Report the evidence map for gate commands, the actual local check output,
whether both CI jobs ran, remote settings verified, and any remaining work.
Use `LOCAL READY` and `REMOTE READY` only when each has been observed.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.