Guide · Agent skill
How to use the Resolve PR Review Comments agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The github-pr-comments skill is a packaged set of instructions and a shell script that tells an AI coding agent how to work through the inline review comments on a GitHub pull request. An agent skill is a folder of instructions (a SKILL.md file plus optional scripts) that the agent loads when a request matches the skill's description. This guide explains what the skill does, how to install it, how to ask for it in plain language, and what you need before it will run.
Resolve PR Review Commentstailcallhq/forgecode · Apache-2.0 Download the skill Quick start
- 1Install prerequisites. Make sure bash, GitHub CLI (gh), jq, a Rust toolchain with cargo, and cargo-nextest are available, and that gh is authenticated with access to the repository.
- 2Download the skill ZIP. Download the ZIP for github-pr-comments from the skill's page.
- 3Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into .claude/skills/ inside your project.
- 4Check the PR. Check out the PR branch, or note the PR number you want to work on.
- 5Ask in plain language. Request something like "resolve the review comments on this PR" so the agent loads the skill.
- 6Review the changes. Let the agent fetch comments, create todos and apply changes, then read the diff carefully.
- 7Confirm verification passed. Check that cargo check and cargo nextest run completed without errors before committing.
What the skill makes the agent do
When you ask the agent to address review feedback on a GitHub pull request, the skill tells it to follow a fixed workflow instead of working it out from scratch each time. The workflow has four steps: fetch every unresolved inline comment, create a todo item per comment, apply each change, and verify the result.
- Fetch: a bundled script, scripts/pr-comments.sh, downloads the inline comments and the diff hunk around each one.
- Track: the agent adds one todo per comment before touching any code, so comments spread across many files are not missed.
- Apply: suggestion blocks are applied verbatim to the highlighted lines; free-form feedback is interpreted in the context of the diff.
- Verify: the agent runs cargo check and cargo nextest run and fixes errors before marking the task complete.
Install the skill
You can install the skill by downloading the ZIP from the skill's page and unzipping it into one of two locations. Use ~/.claude/skills/ to make it available across your projects, or a project's .claude/skills/ folder to share it with that repository.
If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/ so the skill is available in every project.
- Project install: unzip into .claude/skills/ inside the repository.
- Claude apps: upload the ZIP under Settings → Capabilities → Skills.
Trigger the skill with natural requests
The skill's description lists phrases that should activate it. You do not need to name the skill; ask for the outcome you want in ordinary words. The source names these phrases: "resolve review comments", "address PR feedback", "fix PR comments" and "work through review comments".
If you want to target a specific pull request rather than the one for the current branch, include its number in your request. The script accepts an optional PR number.
- "Resolve the review comments on this PR."
- "Address PR feedback on PR 42."
- "Work through the review comments on my current branch."
Run the fetch script yourself
The agent runs the script for you, but you can run it directly to see what it will work from. Omit the PR number to use the current branch's PR. Each block of output contains a File line with the path and line number, a code context section with the diff hunk, and a comment section with the reviewer's message.
- File : file path and line number.
- -- code context --: the diff hunk showing surrounding lines.
- -- comment --: the reviewer's message.
bash .forge/skills/resolve-code/scripts/pr-comments.sh [PR_NUMBER]Prerequisites and what the skill needs
The skill depends on several tools being present on your machine. The script uses GitHub CLI and jq, and the verification step assumes a Rust project. Check these before you ask the agent to start.
- bash
- GitHub CLI (gh), authenticated, with access to the repository
- jq, which the script uses to merge paginated output
- A Rust toolchain with cargo
- cargo-nextest, for the verification step
Safety notes: scripts and network access
This skill runs a shell script and uses the network. The script downloads comments from GitHub through your authenticated gh session, so it can read whatever your account can read in that repository. Review scripts/pr-comments.sh before installing if you are working with sensitive code.
The skill also edits code automatically. Suggestion blocks are applied verbatim, and ambiguous free-form comments are resolved by the agent's own judgment, with the assumption stated. Review the diff the agent produces before you commit or push.
- It runs scripts/pr-comments.sh with bash.
- It contacts GitHub through gh using your credentials.
- It changes files in your working tree without asking for approval on each comment.
- Read the final diff before committing.
Tips for better results
Clear, specific comments produce better changes. Because the agent uses the diff hunk as context, a reviewer comment that refers to a visible line is easier to apply correctly than a general remark about the design.
Because the verification step is hard-coded to cargo check and cargo nextest run, the skill fits Rust projects best. For other languages, expect to edit step 4 of SKILL.md yourself.
- Make sure your local branch is the PR branch before running, so the current-branch lookup finds the right PR.
- Pass the PR number explicitly when you have several open pull requests.
- Ask the agent to list any assumptions it made about ambiguous comments, and check them.
- Keep the work in a clean git state so the diff from the agent is easy to review.
What people use it for
- Working through a long list of reviewer comments on a Rust pull request without missing any.
- Applying exact replacement suggestions that reviewers posted as suggestion blocks.
- Turning free-form review remarks into concrete code changes with a stated assumption for each.
- Running the build and test suite automatically after a batch of review fixes.
Frequently asked questions
Is the github-pr-comments skill free?
The skill is published under the Apache-2.0 licence, which is a permissive open-source licence. Check the licence file in the download for the full terms before you reuse or redistribute it.
What do I need installed to use it?
You need bash, GitHub CLI (gh) authenticated with access to the repository, jq, a Rust toolchain with cargo, and cargo-nextest. The verification step assumes a Rust project.
Where do I put the skill files?
Unzip the download into ~/.claude/skills/ for use across all projects, or into .claude/skills/ inside a specific project. In the Claude apps you can upload it under Settings → Capabilities → Skills.
Can I use this skill on a project that is not written in Rust?
Not as written. Step 4 of SKILL.md hard-codes cargo check and cargo nextest run, so the verification will fail or need editing for other languages.
Does the skill change my code without asking?
Yes. It applies suggestion blocks verbatim and resolves free-form comments using the agent's judgment. Review the resulting diff before committing.
Why does the agent say it cannot find the PR?
The script uses the current branch's PR when you omit the PR number, so check that you are on the PR branch or pass the PR number explicitly. The script also depends on gh being authenticated with access to the repository.
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 Rust developers and maintainers who want an agent to work systematically through a long set of PR review comments, with a todo per comment and a build and test check at the end. It is a poor fit for projects in other languages without editing, or for anyone who does not want an agent editing code without approval on each change.
Download the skill