What does this skill do, and when should you use it?
review-rendercv-pr is an AI code-review skill shipped inside the RenderCV repository. It uses the GitHub CLI to locate and read a pull request, combines that with RenderCV-specific development and testing context documents to understand project conventions, and evaluates the full diff plus the complete contents of every changed file. Its evaluation covers correctness, code conventions, architecture, test coverage, and security, and it also checks CI status. It then posts a structured review on GitHub with one of three verdicts — approve, request changes, or comment — and supports line-level inline comments.
The skill performs the following: locates and reads the PR via gh pr list / gh pr view, including any linked issue; fetches the full diff and changed-file list with gh pr diff; reads the sibling rendercv-development-context and rendercv-testing-context skills to learn project standards; reads the full contents of each changed file on the PR branch (not just the diff); evaluates against correctness, conventions, architecture, testing, and security (injection, path traversal, hardcoded secrets); checks CI with gh pr checks; posts a structured review via gh pr review with sections for Review Summary, Correctness, Code Conventions, Architecture, Testing, and Other Notes; optionally adds line-level inline comments via the GitHub API; and finally reports the verdict, key findings, blocking issues, and a link to the posted review to the user.
- A RenderCV maintainer triaging a backlog of open-source PRs wants a standards-based first pass before human sign-off.
- An external contributor to RenderCV wants to self-check their code against project conventions before requesting review.
- A team wants to confirm before merging whether a PR has failing CI checks or missing test coverage.
- A workflow needs review verdicts (approve / request changes / comment) recorded directly on GitHub as an auditable trail.
How do you install this skill?
- The skill posts comments to GitHub PRs automatically (including approve/request-changes); confirm you accept automated write actions or run under human supervision.
- Fully dependent on GitHub and gh CLI authentication; behavior when unauthenticated or under-privileged is undefined. This was a static review — no commands were executed.
- Applicable only to PR review in rendercv/rendercv; adjust before reusing for other repositories.
- Review depth depends on the companion rendercv-development-context and rendercv-testing-context skills; verify those files exist and are current before use.
- Shell / CLI
- Network access
- Local filesystem
GitHub CLI (gh)RenderCV skill collection (.claude/skills/rendercv-development-context, rendercv-testing-context)
The skill is one of 6 bundled under the RenderCV repository's .claude/skills/ directory; per the README, the skill collection is installed with npx skills add rendercv/rendercv-skill (this documents the whole collection). It also depends on two sibling skills in the same collection: rendercv-development-context and rendercv-testing-context. GitHub CLI (gh) must be installed and authenticated locally.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/rendercv/rendercv.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/.claude/skills/review-rendercv-pr" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Review rendercv/rendercv PR 123 against project standards
In any AI client that supports Agent Skills, trigger the skill and supply a PR number or URL (e.g., "Review rendercv/rendercv PR 123 against project standards"); if no number is given, the skill lists open PRs with gh pr list for you to choose from. When done, it reports the overall verdict, key findings per category, blocking issues, and a link to the posted review.
What are this skill's strengths and limitations?
- Review standards come from RenderCV's own maintained development and testing context documents, keeping feedback aligned with actual project conventions.
- Reads the full contents of changed files on the PR branch, not just the diff, giving review real context.
- Closes the loop by posting the verdict directly to GitHub via gh pr review, including inline line-level comments.
- Explicitly requires security checks and CI status, flagging failing checks as merge blockers.
- Hard-wired to the rendercv/rendercv repository and GitHub CLI; it cannot review other projects as-is.
- Depends on two sibling context skill documents — copying this SKILL.md alone is not enough for it to work.
- Review quality depends on how current those context documents are; SKILL.md itself offers no test or evaluation evidence.
- The README's install command (npx skills add rendercv/rendercv-skill) covers the whole collection; a standalone install path for this skill is not documented in the source.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| RenderCV PR Reviewer this page | 55 · Use with care | ★ 18k | 7mo ago | MIT |
| Ship a PR (e2e repo PR delivery workflow) | 59 · Recommended | ★ 10k | 11d ago | Apache-2.0 |
| Codex Pull Request Editor ✓ OpenAI · Official | 50 · Use with care | ★ 129k | 11d ago | Apache-2.0 |
| Yeet GitHub PR Flow ✓ OpenAI · Official | 41 · Not recommended | ★ 28k | 4mo ago | — |
| RenderCV Typst Package Publisher | 51 · Use with care | ★ 18k | 7mo ago | MIT |
How did FollowSkills review this skill?
The skill uses read-only gh commands (pr list/view/diff/checks) plus one write action (pr review / inline comments); permissions are proportionate, no credential handling or covert exfiltration. Deducted because Step 7 posts reviews directly with the chosen event without requiring user confirmation before approve/request-changes, and rollback/data-flow disclosure is implicit only.
Instructions are self-consistent, commands concrete and copyable, failure path covered via gh pr checks. Static cap of 10 applies: gh authentication availability, --patch behavior, and line-number assumptions of the inline-comment API cannot be verified; failure feedback on abnormal inputs (private repo, missing permissions) is not covered.
Audience and trigger scenario are clear (PR review for the rendercv repo), low false-trigger risk. Deducted for: full dependence on GitHub and gh CLI with unassessed mainland-China reachability, undeclared Chinese-language support, and no stated non-fit boundary for other repositories.
Well-layered flow (identify PR, read context, review), references companion development/testing context skills; MIT license, active repo, and a CI workflow (claude-pr-review.yaml) that actually invokes this skill support maintenance. Deducted for: no versioning/changelog, template parameters (<event>) governed by convention rather than validation, and no known-limitations disclosure.
The workflow covers correctness, conventions, architecture, and testing and produces a directly usable structured review; real marginal value over manual review, corroborated by the CI workflow invoking it. Deducted because the static cap is 7, review quality depends on two referenced context skills not present in evidence, and final output completeness is unverifiable without execution.
An in-repo workflow (.github/workflows/claude-pr-review.yaml) provides third-party usage evidence and all commands are independently reproducible. Deducted because there are no execution results, tests, or eval records specific to this skill, and the key claim (review quality) cannot be independently confirmed; 4 within the static cap of 5.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →