Claude-skill-registry github-prs

Use when responding to PR comments, updating PRs, or handling PR feedback as an AI agent. Ensures proper identification and use of MCP tools.

install

source · Clone the upstream repo

git clone https://github.com/majiayu000/claude-skill-registry

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/github-prs" ~/.claude/skills/majiayu000-claude-skill-registry-github-prs && rm -rf "$T"

manifest: skills/data/github-prs/SKILL.md

GitHub PRs

Overview

Guidelines for AI agents handling pull request interactions. Agents must identify themselves and use MCP tools for GitHub operations.

When to Use

Responding to PR review comments
Replying to PR discussion comments
Updating PR based on feedback
Checking PR status or reviews

Agent Identification

Always identify yourself as an AI agent when commenting on PRs.

Use this signature at the end of PR comments:

— Claude

MCP Tools for PRs

Use the GitHub MCP server tools instead of

gh

CLI when available:

Task	MCP Tool
Read PR details	`mcp__github__pull_request_read` with `method: "get"`
Get PR diff	`mcp__github__pull_request_read` with `method: "get_diff"`
Get PR comments	`mcp__github__pull_request_read` with `method: "get_comments"`
Get review comments	`mcp__github__pull_request_read` with `method: "get_review_comments"`
Get PR checks/status	`mcp__github__pull_request_read` with `method: "get_status"`
Add PR comment	`mcp__github__add_issue_comment` (PRs are issues)
Create review	`mcp__github__pull_request_review_write` with `method: "create"`
Update PR	`mcp__github__update_pull_request`

Reporting PR Checks

When asked about PR checks or status, use

pull_request_read

with

method: "get_status"

and display results in a table:

| Check | Status | Details |
|-------|--------|---------|
| Validate PR Title | :white_check_mark: Passed | - |
| Build | :white_check_mark: Passed | - |
| Lint | :x: Failed | `src/auth.py:42` - Line too long |
| Tests | :hourglass: Pending | Running... |

Status icons:

```
:white_check_mark:
```
- Passed/Success
```
:x:
```
- Failed
```
:hourglass:
```
- Pending/In Progress
```
:warning:
```
- Skipped or neutral

Responding to PR Comments

Read the comment using

pull_request_read

with

method: "get_comments"

"get_review_comments"

Understand the feedback - what change is being requested?
Make the fix if it's a code change request
Reply to acknowledge using
```
add_issue_comment
```
Include agent signature in your response

Example Response

Thanks for the feedback! I've updated the code to address your concern.

Changes made:
- Fixed the validation logic in `auth.py:45`
- Added the missing error handling

— Claude

Handling Review Feedback

When a PR review requests changes:

Read review comments with
```
pull_request_read
```
method
```
"get_review_comments"
```
Address each comment with code fixes
Commit and push fixes
Reply to each review thread if possible, or add a summary comment

Summary Comment Template

## Addressed Review Feedback

| Comment | Resolution |
|---------|------------|
| [Brief description] | [What was fixed] |
| [Brief description] | [What was fixed] |

All feedback has been addressed. Ready for re-review.

— Claude

Key Principles

Always identify as AI - transparency about agent-generated responses
Use MCP tools - prefer MCP GitHub tools over CLI when available
Acknowledge feedback - always respond to comments, even if no code change needed
Be concise - keep responses focused and actionable
Reference changes - point to specific files/lines when discussing fixes