Guide · Agent skill
How to use the PR Report Writer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The pr-report skill is an agent skill from the paperclipai/paperclip project that makes an AI agent write a structured, maintainer-style review of a pull request, branch, or large code contribution. It is aimed at open-source maintainers, engineers reading an unfamiliar change, and teams deciding whether to merge a significant feature. After this guide you will know what a skill is, how to install pr-report, how to ask for a report in plain language, and where its limits are.
PR Report Writerpaperclipai/paperclip · MIT Download the skill Quick start
- 1Download the ZIP. Get the pr-report skill ZIP from the page where you found this guide.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use or into the project's .claude/skills/ folder for one repository.
- 3Or upload in the Claude apps. If you use the Claude apps, go to Settings → Capabilities → Skills and upload the ZIP.
- 4Make a natural request. Ask for a deep review, explanation, or merge recommendation for a named PR, branch, or folder.
- 5Choose the output. Ask for an HTML webpage or a Markdown file, and name a folder if you have a preference.
- 6Check the report. Open the saved report and verify its findings against the actual code before acting on them.
What is an agent skill and what does pr-report do?
An agent skill is a folder of instructions and supporting files that an AI agent loads when a task matches the skill's description. The pr-report skill is defined in a SKILL.md file. Its description says it produces a maintainer-grade PR or contribution report as HTML or Markdown, for requests to deeply review a PR, explain a design, compare systems, or prepare a merge recommendation.
The skill asks the agent to understand what the change is trying to do before judging it, to explain the system as built rather than just the diff, and to separate architectural problems from product-scope objections. Findings are ordered by severity and the report ends with a concrete recommendation.
- Explains what the change adds and how the system works, with a tutorial-style section for large contributions
- Lists risks and strengths, prioritised by severity, with file references where possible
- Compares the design to similar open-source systems when relevant
- Ends with one recommendation: merge as-is, merge after specific redesign, salvage specific pieces, or keep as design research
- Outputs a standalone HTML page or a Markdown file, using bundled style guidance and an HTML starter template
Install the pr-report skill
Download the skill ZIP from the page where you found this guide. Unzip it into one of two locations, depending on how you want to use it. A personal install makes the skill available across your projects. A project install keeps it only in that repository.
If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Trigger the skill with natural requests
You do not need a special command. The agent loads the skill when your request matches its description. The SKILL.md lists requests that fit, such as asking to review a PR deeply, explain a contribution, make a report or webpage for a PR, compare a design to similar systems, or whether you should merge.
Be specific about the target (a PR number, a branch, or a folder) and the output you want (a webpage or a Markdown file). Mention the output folder if you have a preference.
- Review this PR deeply and tell me the main risks.
- Explain this contribution to me as a tutorial before I review it.
- Make me a webpage report for this branch.
- Compare this design to similar systems and say what to keep.
- Should I merge this? Write the report as Markdown in report/.
Where the reports are saved
The skill names two common output types. A standalone HTML report is written under tmp/reports/. A Markdown report is written to report/ or another folder you request. The agent also gives a short maintainer summary in chat, which the skill asks to be shorter than the report itself.
For HTML output, the skill suggests starting from the bundled starter file and replacing its placeholder content, and following the style guide for typography and layout.
- HTML report: tmp/reports/...
- Markdown report: report/ or your chosen folder
- Chat: a short summary with the location, overall call, top reasons, and whether tests were skipped
What the skill needs and its safety notes
The listing states that the skill requires nothing beyond the agent itself, runs no scripts, and uses no network access. Its bundled files are references/style-guide.md, assets/html-report-starter.html, and SKILL.md.
The skill does not run or test the code. The SKILL.md asks the agent to verify findings against the actual code and to say explicitly when tests were not run. Treat each report as a set of claims to check against the real repository before you act on them.
- Requires: nothing listed
- Runs scripts: no
- Uses network: no
- Does not execute or test the contribution
- Check findings against the real code before merging or rejecting
Tips for better reports
Work from a local checkout when you can. The skill's workflow asks for local code rather than only the GitHub PR page, so give the agent a branch or worktree to read. Include any design docs or PR text that explain intent, because the skill looks for contributor intent there.
Use this skill for significant changes. For a quick typo or small bug fix, the multi-section workflow is more than you need.
- Point the agent at a local branch or worktree
- Mention relevant specs or design docs
- Ask for a specific output format and folder
- Use it for large or complex changes, not small fixes
What people use it for
- A maintainer evaluating a large pull request that adds a new subsystem before deciding whether to merge it
- An engineer who needs a readable walkthrough of an unfamiliar contribution before doing their own review
- A team preparing a scope or redesign decision for a significant feature, with a written recommendation to discuss
- Comparing a proposed design with similar open-source projects to extract lessons about extension boundaries and trust models
Frequently asked questions
How do I install the pr-report skill?
Download the skill ZIP from the page where you found this guide and unzip it into ~/.claude/skills/ for personal use, or into the project's .claude/skills/ folder for a single repository. In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.
Is the pr-report skill free?
The skill is listed under the MIT licence. Check the licence file in the source project for the full terms before reusing or redistributing it.
Does pr-report need any other tools or scripts?
The listing says it requires nothing beyond what is listed, runs no scripts, and uses no network access. The skill's files are SKILL.md, a style guide, and an HTML starter template.
Will pr-report run or test the code in my pull request?
No. The skill asks the agent to verify findings against the code, but it does not include test execution scripts. Check the report against the real code, and if tests were not run, the report should say so.
What is the difference between the HTML and Markdown output?
HTML produces a standalone webpage with clear sections, saved under tmp/reports/. Markdown produces a plain text report saved to report/ or a folder you name. Choose HTML for easy reading and Markdown for pasting into issues or docs.
When should I not use pr-report?
Skip it for quick typo fixes or small bug fixes, since the workflow is built for deep, multi-section reports and would be overkill. For a small change, a normal code review is faster.
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
Use pr-report if you maintain an open-source project or review large, complex contributions and want a structured, readable report with a clear merge recommendation. Pick something else if you need the code run or tested automatically, or if you only review small changes, since this skill only analyses and writes up what it reads.
Download the skill