Skip to content
Get packs

Guide · Agent skill

How to use the ClawHub PR Maintainer agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The clawhub-pr-maintainer skill is an agent skill for maintainers of the openclaw/clawhub repository. When it is installed, the agent follows a strict maintainer checklist for GitHub issues and pull requests instead of its usual general behavior. This guide explains what the skill does, how to install it, how to trigger it with natural requests, and what it needs and where to be careful.

ClawHub PR Maintaineropenclaw/clawhub · MIT

Quick start

  1. 1Install the gh CLI and log in. Install the GitHub CLI and authenticate it with an account that has access to openclaw/clawhub. The skill's read-only commands depend on this.
  2. 2Download and place the skill. Download the skill ZIP from this page, unzip it, and put the folder in ~/.claude/skills/ or a project's .claude/skills/ folder. Alternatively, upload the ZIP under Settings → Capabilities → Skills.
  3. 3Confirm the skill is loaded. Start a session and ask the agent to review a specific openclaw/clawhub PR or issue by number. Check that it reads live state with gh before giving any verdict.
  4. 4Ask for a review with a clear scope. Name the repository and item number, and say whether you want a review, a triage or a handoff note. The skill's review output format applies to PR reviews.
  5. 5Set up UI proof only when needed. For visual changes, prepare Bun, the repo's proof:ui script, Playwright and a local ClawHub instance with Convex state. Use the local runner only with localhost or loopback URLs.

What is an agent skill, and what does this one do?

An agent skill is a packaged set of instructions that an AI agent loads when a task matches the skill's description. This skill's description says it applies when reviewing, triaging, validating, or discussing ClawHub GitHub issues or pull requests, including author context, CI, UI proof, evidence, labels, close decisions, and maintainer handoff.

The skill is scoped to maintainer workflow. It is explicitly not for ordinary feature implementation work.

  • Reads live GitHub state with the gh command line tool before commenting, labeling, closing or recommending a merge.
  • Checks evidence such as screenshots, tests and root-cause analysis before recommending merge or close.
  • Structures PR review output with size stats, provenance, a best-fix verdict and remaining uncertainty.
  • Can generate UI proof by writing Playwright scenarios and running a proof runner against a local ClawHub instance.

How do you install the clawhub-pr-maintainer skill?

Download the skill ZIP from this page and unzip it. Put the resulting skill folder in ~/.claude/skills/ to make it available to you, or in a project's .claude/skills/ folder to make it available only within that project.

You can also upload the skill in the Claude apps. Go to Settings, then Capabilities, then Skills, and add the ZIP there.

  • Personal install: ~/.claude/skills/
  • Project install: .claude/skills/ inside the project folder
  • Claude apps: Settings → Capabilities → Skills

What does the skill need before it works?

The skill depends on several tools. Only the GitHub CLI is needed for read-only review. The UI proof features need more setup. The skill runs scripts and uses the network, so review what it will run before you install it.

Before you rely on it, confirm you have each item below.

  • GitHub CLI (gh) installed and authenticated with access to openclaw/clawhub
  • Bun runtime and the repo's proof:ui script, for UI proof generation
  • Playwright, for local UI proof runs
  • A local ClawHub instance with Convex state, for the local proof fallback
gh pr view <number> --repo openclaw/clawhub --json title,body,author,labels,comments,files,commits,statusCheckRollup,reviewDecision,url,additions,deletions,changedFiles

How do you trigger the skill with natural requests?

You do not need special syntax. Ask the agent in plain language about a ClawHub issue or pull request, and the skill's description should match. Name the repository and the item number so the agent can read the live state.

Requests that match the skill's described scope include reviewing a PR, triaging an issue, checking whether a bug fix has real evidence, or writing a maintainer handoff note.

  • Review openclaw/clawhub PR 123 and tell me whether the fix has evidence.
  • Triage issue 456 in openclaw/clawhub and suggest labels.
  • Check the CI status and UI proof for PR 789 before I decide to merge.
  • Write a maintainer handoff for PR 101, covering the open uncertainty.
gh issue view <number> --repo openclaw/clawhub --json title,body,author,labels,comments,state,url

What does a PR review from this skill look like?

The skill sets a fixed output shape. A review starts with one to three plain sentences on what the change does and why it matters, then shows size as LOC: +x/-y (N files). Findings come first. If there are none, the review says No blocking findings or No findings.

For bug fixes, the review must state a best-fix verdict, alternatives considered, the code read, and remaining uncertainty. It must also answer whether the fix is the best fix or only a plausible one. Merge is not recommended for a bug fix based only on issue text, PR text or AI rationale.

  • Best-fix verdict: best, acceptable mitigation, wrong layer, too narrow or too broad
  • Alternatives considered: one to three concrete alternatives and why they were rejected
  • Code read: the main files or contracts checked
  • Remaining uncertainty: what was not proven

What safety points should you know before running it?

Close, label and comment actions affect a public project. The skill says to verify live state before taking them. Anyone without GitHub write access should not run close or label actions, and the skill's guidance is aimed at people who have that access.

The skill runs scripts and uses the network. Its UI proof runner accepts only localhost or loopback URLs. Keep proof screenshots and videos in .artifacts/ until publishing, and never commit proof artifacts. The skill also tells the agent not to rerun proof:ui just to inspect evidence that already exists.

  • Verify live GitHub state before commenting, labeling, closing or recommending merge.
  • Treat author identity as a triage signal only, never as proof by itself.
  • Do not merge based only on PR prose, AI rationale or green CI when the changed behavior has not been exercised.
  • Never commit proof artifacts.

What people use it for

  • Triaging incoming issues on openclaw/clawhub and recording labels and open questions before a maintainer decides.
  • Writing a consistent, evidence-based review of a bug-fix PR with a best-fix verdict and listed remaining uncertainty.
  • Checking whether a contributor's screenshots, logs or tests actually prove the claimed behavior.
  • Generating before-and-after UI proof for a visual change by running a Playwright scenario against main and a candidate checkout.

Frequently asked questions

Is the ClawHub PR Maintainer skill free to use?

The skill is listed under the MIT licence. The source does not state any price, so there is no paid tier described for it. You still need your own GitHub access and any tools it depends on, such as gh, Bun and Playwright.

What do I need to install the clawhub-pr-maintainer skill?

You need the skill ZIP from this page, placed in ~/.claude/skills/ or a project's .claude/skills/ folder, or uploaded under Settings → Capabilities → Skills. For read-only review, you also need the GitHub CLI (gh) authenticated with access to openclaw/clawhub.

Can I use this skill on repositories other than openclaw/clawhub?

No. The source says the commands are hard-coded to openclaw/clawhub and its scripts, so the skill is not designed for other repositories.

Does the skill merge or close pull requests for me?

The skill's guidance is to verify live state before commenting, labeling, closing or recommending merge. It is designed to structure a review and recommend a decision. People with GitHub write access should review the output and make close or label actions themselves.

Why does the skill say it needs a local ClawHub instance?

The local UI proof fallback runs a Playwright scenario against a real local ClawHub instance with Convex state. The proof runner accepts only localhost or loopback URLs, so the instance must run locally.

Is this skill good for writing ordinary feature code?

No. The skill states it is for maintainer-facing GitHub workflow and not for ordinary implementation work. Use a general coding setup for feature development.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits maintainers of openclaw/clawhub who triage PRs and issues and want consistent, evidence-based reviews, and contributors who want to know the bar maintainers will apply. It is a poor fit for work on other repositories, for ordinary feature implementation, or for anyone without GitHub write access who would run close or label actions on a public project.

Download the skill