RenderCV Issue Triage
Analyzes newly opened GitHub issues on rendercv/rendercv, posts a comment with findings and an action plan, and offers to open a PR.
The skill only reads public issues and posts comments via gh CLI; scope matches the paired workflow permissions (contents: read / issues: write), so least privilege largely holds; a PR is only offered after a maintainer reply, with no automatic external code writes. Deducted: posting a public GitHub comment is an externally visible write with no user-confirmation or rollback mechanism in either the skill or the workflow, and the data flow (issue content sent to the Anthropic API) is not disclosed.
The three steps (read issue, check duplicates, post comment) are self-consistent with concrete, copyable commands, and it degrades gracefully ('ask clarifying questions instead of guessing'). Deducted: no failure feedback for unauthenticated gh or network errors, duplicate detection depends on keyword quality with no verification, step 2 depends on another skill file's availability, and there is no test evidence for this skill's key paths.
Trigger scenario is clear — automated triage of newly opened issues — and the workflow binding (issues: opened) gives precise triggering with low false-positive risk; comment structure and non-fit cases (duplicate, out-of-scope, wontfix) are described. Deducted: non-fit boundaries are not fully explicit, and core function depends on the GitHub and Anthropic APIs, raising mainland-China reachability concerns.
The skill file is well structured with reasonable progressive disclosure; the repo has an MIT license, active workflows, and a versioned main project (rendercv 2.8), so the maintenance path is visible. Deducted: SKILL.md itself has no version or changelog, hidden assumptions (authenticated gh, configured ANTHROPIC_API_KEY) are not stated in the skill, and the publisher is unverified.
The task (a valuable triage comment) is well defined; the five-part comment structure should yield directly usable output, and the marginal value lies in replacing manual codebase reading to locate issues. Deducted: static review cannot verify actual comment quality, output depends entirely on model judgment, and the repository's promptfoo evals cover a different skill path, not this one.
The commands and the paired claude-issue-triage.yaml workflow are auditable primary material with one-to-one correspondence to the skill. Deducted: no third-party execution evidence, no test or eval reproduction covering this skill path, and conclusions are mostly static inference.
- The skill automatically posts comments to public GitHub issues (an external write) with no human confirmation or rollback; validate behavior in a fork or test repo first and be prepared to delete or edit comments manually.
- Core function depends on the GitHub API and Anthropic API, both of which may be unreachable from mainland-China networks, potentially making the skill unusable there; a working network path and ANTHROPIC_API_KEY are required.
- Publisher identity is unverified in the FollowSkills registry; this is a static source review with no commands executed, so actual comment quality and duplicate-detection accuracy are unverified.
- The skill assumes gh CLI is installed and authenticated and that repository secrets are configured; these prerequisites are not stated in SKILL.md.
What does this skill do, and when should you use it?
One of six skills bundled in the rendercv/rendercv repository, this skill does a single job: triage newly opened issues on that repo. It reads the issue with all comments, classifies it (bug, feature request, or question), searches for duplicates, and uses a sibling development-context skill to understand the relevant code. It then posts a five-part comment: restating the problem, file-level analysis, concrete fix steps, pitfalls to avoid, and an invitation to reply @claude for a PR. It suits RenderCV maintainers and developers who want the same disciplined triage workflow for their own repos.
Reads the full issue and comments via gh issue view <number> --repo rendercv/rendercv --comments; checks for duplicates with gh issue list --search ...; consults .claude/skills/rendercv-development-context/SKILL.md for architecture and explores related source files and tests; posts a structured comment via gh issue comment containing an understanding restatement, file-level analysis, a proposed approach with specific files and functions, things to avoid, and a PR offer. It deliberately does not label or assign the issue and never promises timelines.
- A RenderCV maintainer wants to offload first-pass replies to newly opened issues.
- An open-source contributor with a high-traffic issue queue needs duplicates, incomplete reports, and out-of-scope requests flagged quickly.
- A team wants standardized triage comments: restate the problem, then give a per-file fix path.
- A developer modeling an issue-triage skill for their own repo can reuse its explicit comment template and boundary rules.
What are this skill's strengths and limitations?
- The comment structure is explicit (understanding, analysis, approach, pitfalls, offer), so output is predictable and easy to review.
- Built-in duplicate checking plus clear boundaries: no labels, no assignments, no timeline promises, no unrequested close suggestions.
- Pairs with a repository development-context skill, so analysis references real files and functions rather than generic advice.
- Tightly coupled to the rendercv/rendercv repo; adapting it to another project requires rewriting repo references and the context skill.
- Requires an authenticated GitHub CLI and posts publicly on real issues, so mistakes are visible.
- No test or evaluation results are provided for this skill in the source material.
How do you install this skill?
The skill ships inside the rendercv/rendercv repository at .claude/skills/triage-rendercv-issue/SKILL.md. The README's install command for the skill collection is npx skills add rendercv/rendercv-skill, which works with any agent supporting the skills standard; placing the skill file under your .claude/skills/ directory makes it discoverable by Claude Code. Note: the README describes the whole collection; per-skill standalone install steps are not documented in the source.
How do you use this skill?
In an environment with an authenticated GitHub CLI (gh), trigger it with a prompt like "Triage issue #123 on rendercv/rendercv". It will read the issue, search for duplicates, analyze the relevant code using the development context, and post a five-part comment ending with the offer: "Reply @claude followed by your instructions if you'd like me to open a PR for this." If the issue is unclear, it asks clarifying questions instead of guessing.