From devkit
Review a GitHub pull request via the gh CLI and post structured review comments. Use when the user says "review this PR", "look at PR
How this skill is triggered — by the user, by Claude, or both
Slash command
/devkit:review-prThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You are conducting a pull request review using the `gh` CLI. Be direct, specific, and prioritise ruthlessly. Always work from the actual diff — never from assumptions.
You are conducting a pull request review using the gh CLI. Be direct, specific, and prioritise ruthlessly. Always work from the actual diff — never from assumptions.
gh pr diff <number>gh pr view <number> — read the description to understand stated scope.Evaluate the diff across these dimensions:
Correctness & Risk
Design & Architecture
Code Quality
Low-Hanging Fruit
Always end every comment body — review summary, top-level comment, or inline comment — with a newline followed by — Claude so the PR author can tell the feedback came from an AI assistant rather than a human reviewer.
Do not post review comments directly. Drafts are reviewed by the user first.
After reading the diff and forming findings, present a single preview in the chat covering everything you intend to post. For each finding show:
— Claude signature, rendered in a fenced block.Use a format like:
### Inline — src/foo.ts:42
**Category:** correctness
**Draft:**
> <draft comment text>
>
> — Claude
Include the end-of-review summary comment in the preview too.
After showing all drafts, stop and ask the user how to proceed. Offer choices like: post all as-is, edit a specific draft, drop a finding, or cancel. Only post after explicit approval, and post exactly the approved text — do not silently re-edit between preview and posting.
Post top-level / summary comments with:
gh pr review <number> --comment --body "$(cat <<'EOF'
<comment text>
— Claude
EOF
)"
For inline comments on specific lines:
gh api repos/{owner}/{repo}/pulls/<number>/comments \
--method POST \
--field body="$(cat <<'EOF'
<comment text>
— Claude
EOF
)" \
--field commit_id="<sha>" \
--field path="<file>" \
--field line=<line>
End the review with a short summary comment (max 3 lines) covering overall impression, top concern, and recommended next step. Sign it the same way.
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub markusjohansen/mj-plugins --plugin devkit