Skip to content
Skills Library
Programming · Automation

install

/install

Install the Gatehand autonomous coding agents (coder, fixer, Claude review, merge gate) into the current GitHub repository. Use when the user asks to install, set up or add the autonomous agents / Gatehand to a repo.

Repository
gatehand ↗
Licence
MIT
Last repository change
12.11.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
1
Deprecation

No signals detected. Repository changed on 12.11.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Install the Gatehand autonomous coding agents (coder, fixer, Claude review, merge gate) into the current GitHub repository. Use when the user asks to install, set up or add the autonomous agents / Gatehand to a repo.

Original author description

Install the Gatehand autonomous coding agents (coder, fixer, Claude review, merge gate) into the current GitHub repository. Use when the user asks to install, set up or add the autonomous agents / Gatehand to a repo.

How it works

  1. 1. Check the ground
  2. 2. Learn the repository
  3. 3. Decide with the user
  4. 4. Render
  5. 5. The steps only the user can do
Original SKILL.md headings · 92 lines, 764 words
SKILL.md preview
---
name: install
description: Install the Gatehand autonomous coding agents (coder, fixer, Claude review, merge gate) into the current GitHub repository. Use when the user asks to install, set up or add the autonomous agents / Gatehand to a repo.
---

# Install Gatehand into this repository

The package root is two directories above this file (this file is `<root>/skills/install/SKILL.md`).
`<root>/bin/render.sh` does the file work; your job is to fill in the repo's settings well, and to
walk the user through the steps only they can do. Read `<root>/docs/README.md` first: it is the
runbook that will be installed, and it explains every setting below.

Never push, open a PR, create a GitHub App, set a secret or variable, or change repository settings
without the user's explicit go-ahead for that step. Prefer giving them the exact command.

## 1. Check the ground

- The current directory is the repository's root, a git checkout with a clean working tree, and
  `gh auth status` works. `bash --version` is 4 or later (macOS: `brew install bash`), `jq` exists.
- `gh repo view --json nameWithOwner,defaultBranchRef,visibility,isFork`. Refuse a fork: the agents
  need secrets that forks do not get.
- If `.github/agents/install.lock.json` exists, this is an update: use the update skill instead.

## 2. Learn the repository

Find, without changing anything:
- The CI workflow that runs on pull requests: its file name and its `name:`, and the job names that
  should gate a merge (these become `requiredChecks`; a job's check name is its `name:`, else its id).
- The language(s), package manager and lockfile, and the exact commands CI runs to build, lint,
  typecheck and test, per area of the repo.
- What CI installs first (runtimes, system packages).
- `CLAUDE.md` or `AGENTS.md` and other contributor docs: standing rules worth binding the agent to.
- Release or deploy tooling (`deploy/`, `scripts/release*`, `Dockerfile`s CI pushes): candidates for
  protected paths.
- Existing rulesets: `gh api repos/<o>/<r>/rules/branches/<default>`.

## 3. Decide with the user

Ask (AskUserQuestion), recommending defaults:
- The GitHub App's name (its slug becomes `botLogin`, `<slug>[bot]`). Default: `<repo>-gatehand`.
- The owner login whose inbox commands count. Default: `gh api user --jq .login`.
- Claude auth: `oauth` (a `claude setup-token` subscription token) or `api-key` (an Anthropic
  Console key, which can be spend-capped; recommended for teams).
- Runners: GitHub-hosted (default) or a self-hosted pool (then read
  `<root>/addons/hardened-runner/README.md` with them first).
- Which CI checks must pass before an agent PR merges.

## 4. Render

On a new branch `chore/install-gatehand`:

```bash
bash <root>/bin/render.sh .
```

That writes the package files and, once, the files that are the repo's own under `.github/agents/`.
Then edit those to fit what you found:
- `policy.json`: `botLogin`, `baseBranch`, `ci.workflowName`, `ci.workflowFile`, `requiredChecks`,
  `runners`, `claude.auth`, and `tools.allow` with one `Bash(<command prefix>:*)` per check command
  in `project.md` (for example `Bash(npm test:*)`, `Bash(go:*)`). Keep the defaults for the limits.
  If a check command matches a `tools.deny` entry or the built-in deny list (git push, gh, curl,
  docker, sudo…), it cannot be allowed: say so.
- `project.md`: the standing rules and, under "Checks", the commands per area. Concrete and short.
- `protected-paths.txt`: add the release and deploy tooling you found as normal (label-unlockable)
  lines. Never remove a `!` line.
- `system/action.yml` (sudo, fixed packages only), `toolchain/action.yml` (setup-* actions with
  every cache off: setup-go takes `cache: false`; setup-node and setup-python take a package
  manager's name in `cache:` and reject `false`, so leave it out, and on setup-node v5 or later set
  `package-manager-cache: false`),
  `dependencies/action.yml` (lockfile-exact install): mirror what CI does.

Re-render (`bash <root>/bin/render.sh .`) and confirm `bash <root>/bin/render.sh --check .` passes.
If the repo lints workflows with actionlint, add the ignore rules from
`<root>/tests/lint/actionlint.yaml` (and any self-hosted labels) to its `.github/actionlint.yaml`.

Show the user a summary of the diff and the settings you chose, then commit on their go-ahead
(`chore: install Gatehand <version>`) and, on a second go-ahead, push and open the PR.

## 5. The steps only the user can do

Give them these, in order, as commands where possible (the runbook's "Setup" has the details):

1. Create the GitHub App: the manifest is `<root>/docs/app-manifest.json` (set `name` to the chosen
   slug). Permissions: Contents, Issues, Pull requests, Actions read and write; Checks, Commit
   statuses, Metadata read; **no Workflows**. Install it on this repository only.
2. Install the Claude GitHub App on the repository (https://github.com/apps/claude).
3. Secrets and variables: `AGENT_APP_PRIVATE_KEY`, `AGENT_APP_CLIENT_ID`, `AGENT_BOT_ID`,
   `AGENT_OWNER_LOGIN`, and `CLAUDE_CODE_OAUTH_TOKEN` or `ANTHROPIC_API_KEY`.
4. Actions defaults: read-only token, no PR approval by Actions.
5. After the install PR merges: `bash .github/scripts/agent-labels.sh`, then
   `git push origin origin/<default>:refs/heads/agent-runner`.
6. Run the doctor skill. Then `gh variable set AGENTS_ENABLED --body true`, and queue one small
   issue with an acceptance section: `gh workflow run agent-coder.yml --ref agent-runner -f issue=<n>`.
7. After the first agent PR: pin the inbox issue and make `agent pr guard` a required check. Leave
   `AGENT_AUTO_MERGE` off until a few merge-gate dry runs read right.

Versions

RepositoryScoreLicence
This page · codifydoo/gatehand100MIT
rossoctl/context-guru100Apache-2.0
jeremylongshore/tons-of-skills-marketplace100MIT
terrylica/cc-skills100MIT
agile-4/opem-agent-plugin100MIT
artk0de/tearags-mcp100MIT
artk0de/tearags-mcp100MIT
brainweb/payload-mcp-oauth100MIT
x-cmd/x-cmd100Apache-2.0

Database snapshot: 13.11.2026 · package version e0289f48e349 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 96update
    Update the Gatehand files in the current repository to the installed package version, or re-render them after a policy.json change. Use when the user asks to update or upgrade Gatehand, or changed .github/agents/policy.json.
  • 100agent-readiness
    Score how ready a repository is for autonomous coding agents (5 levels, 9 pillars, 93 criteria) and render an HTML dashboard plus a markdown summary. Use when the user asks for an agent readiness report, score or audit, asks whether a codebase is "agent-ready" or why agents…
  • 100agentic-setup
    Sets up or upgrades a repository for AI coding agents (Claude Code, Codex, Cursor) with current best practices — AGENTS.md as an index, topic docs, a decision log, shared skills, PR-template hooks, CI with required checks, branch protection and deploy-from-main. Use when the…
  • 100dev-swarm-install-ai-code-agent
    Install AI code agent CLI tools including claude-code, gemini-cli, codex, and github copilot-cli. Use when setting up AI coding assistants or when the user asks to install an AI code agent.
  • 100agents-md
    Creates and maintains concise AGENTS.md and CLAUDE.md project instruction files. Use when asked to create AGENTS.md, update AGENTS.md, maintain agent docs, set up CLAUDE.md, document repository agent conventions, or keep coding-agent instructions minimal and reference-backed.
  • 100install-agents
    Use when the user wants to install or update the agents of this repository into Claude Code, Codex, Gemini CLI or Copilot CLI (user-level or in one project).

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.