github-pr-review
Format and publish completed review findings as GitHub PR comments or submitted reviews.
Skill by tobywise
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/tobywise-github-pr-review
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/github-pr-review
What's inside
- Skill name
github-pr-review- Description
- Format and publish completed review findings as GitHub PR comments or submitted reviews. Use only when the user requests posting to GitHub. For assessing a PR, including requests to review and post, use review-implementation first. Do not select this skill for reviews or drafts requested only in chat.
- License
- MIT
- Instructions
- 79 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Post a GitHub PR review Publish a completed assessment with clear evidence links and ratings. ## Use the completed review - For a request to review and post, load and complete `review-implementation` first. This skill handles formatting and submission after that assessment. - If the user already supplies a completed review to post, use it. If an assessment is missing or needs further review, use `review-implementation` before submission; do not conduct a separate review under this skill. - Preserve finding IDs, blocking decisions, classifications, scientific tiers, severity, likelihood, rating reasons, and stated uncertainty. Keep possible risks distinct from observed defects. Do not invent fixes for risk candidates. - Identify the intended PR and confirm the reviewed commit still matches its head. If it has changed, use `review-implementation` to assess the new changes before posting. Check existing comments when useful to avoid duplicates. ## Reference every finding Every finding and risk candidate needs a clickable link to its supporting evidence. An inline GitHub review comment anchored to the relevant line counts as its reference. For findings in the review body, prefer a GitHub permalink to the relevant lines at a specific commit. Link further locations when needed to understand the issue. Check that each link opens the intended evidence before posting. A plain `path:line` is not enough. If a required reference cannot be located, report the missing reference before submission. Do not silently drop or reclassify a finding to make the review publishable. ## Preserve the ratings Include the severity, likelihood, and evidence-based reason for each from the completed review, for every finding and risk candidate. Preserve `UNKNOWN` and the named missing fact when a rating cannot be supported. If these fields are missing, complete them through `review-implementation` before posting. Formatting must not change ratings or decide anew whether a finding blocks. ## Format the review State the reviewed commit and diff scope in the summary, including any limits on coverage. Keep it brief and put the most consequential findings first. Use this shape when there are findings: ```markdown ### Findings 1. **Blocking: Short description** — [relevant code](GITHUB_PERMALINK) - **Severity: 3 MAJOR** — Explain the consequence. - **Likelihood: 2 POSSIBLE** — Explain the trigger. - **Impact and fix:** Explain what happens and the smallest practical fix. ### Verification State which checks ran and their results. Say "Not run" if no checks ran. ``` Omit empty sections. If there are no findings, state what was inspected and any limits on that conclusion. Do not claim the PR is correct solely because no issue was found. Do not claim a check ran unless it did. ## Post to GitHub - Post only when the user requests posting; that request authorizes the submission. If the user changes the request to draft only, show the prepared text without submitting it. - Honor the requested form of submission. For a plain PR comment, use `gh pr comment`. For a formal review, use `gh pr review` with the event supported by the assessment: `REQUEST_CHANGES` for blockers, `COMMENT` for nonblocking or unresolved feedback, and `APPROVE` only when approval is intended and supported. Explain an `ESCALATE` assessment in the body; it is not a GitHub review event. Do not infer approval from a limited review. - Write the body to a file and pass it to the chosen command with `--body-file` to preserve Markdown and newlines. Use inline comments for findings when appropriate. - After posting, report the review decision and link to the posted review.
From tobywise/agent-skills, MIT.
More from tobywise/agent-skills
- 0.570score
review-implementation
tobywise
Review pull requests, local implementation changes, or fixes against the applicable requirements.
SkillAFreshness: AClaude Code +9 - 0.570score
planning
tobywise
The shared format and rules for implementation plans.
SkillAFreshness: AClaude Code +9 - 0.566score
run-checks
tobywise
How to run tests, builds, type checks, installs, and other verification commands on this machine — locally, or on remote Crabbox workers when they are…
SkillAFreshness: AClaude Code +9 - 0.565score
orchestrate
tobywise
A flexible guide for carrying out a plan from the main session by delegating implementation and review to agents and running verification yourself.
SkillAFreshness: AClaude Code +9 - 0.565score
plan-review
tobywise
Present a proposed plan with several real decisions in a local HTML page, collect answers, and revise it in rounds before writing the final plan.
SkillAFreshness: AClaude Code +9 - 0.565score
pr-description
tobywise
Draft a reviewer-focused pull request description from the current branch changes.
SkillAFreshness: AClaude Code +9 - 0.565score
create-scientific-rework-plan
tobywise
Audit an existing scientific codebase and plan a safe single-use rework.
SkillAFreshness: AClaude Code +9 - 0.565score
prototype
tobywise
Build and optionally run one rapid, exploratory, single-use scientific analysis prototype, without production engineering ceremony.
SkillAFreshness: AClaude Code +9 - 0.561score
scientific-risk-review
tobywise
Independently rate the severity and likelihood of proposed blockers or possible scientific risks, judge whether they should block, and whether a blocker is…
SkillAFreshness: AClaude Code +9 - 0.559score
create-multiple-plans
tobywise
Create approved, ordered implementation plans for independently useful vertical slices.
SkillAFreshness: AClaude Code +9 - 0.559score
create-plan
tobywise
Create one implementation-ready Markdown plan from a request, or a revision of an existing plan.
SkillAFreshness: AClaude Code +9 - 0.558score
scientific-analysis
tobywise
Use ONLY when planning, implementing, reviewing, or orchestrating a plan whose review standard is SCIENTIFIC_ANALYSIS.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Sep 29, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/tobywise/agent-skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from tobywise
- 0.570score
planning
tobywise
The shared format and rules for implementation plans.
SkillAFreshness: AClaude Code +9 - 0.570score
review-implementation
tobywise
Review pull requests, local implementation changes, or fixes against the applicable requirements.
SkillAFreshness: AClaude Code +9 - 0.566score
run-checks
tobywise
How to run tests, builds, type checks, installs, and other verification commands on this machine — locally, or on remote Crabbox workers when they are…
SkillAFreshness: AClaude Code +9 - 0.565score
create-scientific-rework-plan
tobywise
Audit an existing scientific codebase and plan a safe single-use rework.
SkillAFreshness: AClaude Code +9 - 0.565score
plan-review
tobywise
Present a proposed plan with several real decisions in a local HTML page, collect answers, and revise it in rounds before writing the final plan.
SkillAFreshness: AClaude Code +9 - 0.565score
orchestrate
tobywise
A flexible guide for carrying out a plan from the main session by delegating implementation and review to agents and running verification yourself.
SkillAFreshness: AClaude Code +9
Cite this page
github-pr-review skill. Agent Mart. https://agentmart.dev/skills/tobywise-github-pr-review (last updated 2026-10-24)