skills/fix-issue/SKILL.md
Plans, implements, reviews, and ships a fix for a GitHub issue. Use when fixing an issue, resolving a bug report, implementing a feature request, or closing a ticket. Handles fork workflows, upstream detection, branch creation, and PR submission.
npx skillsauth add philoserf/claude-code-setup skills/fix-issueInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Fix GitHub Issue #$ARGUMENTS.
gh issue view $ARGUMENTS --repo <canonical>
Understand the full context: problem description, acceptance criteria, linked PRs, and discussion. Follow linked issues, referenced PRs, and external documentation to build complete understanding.
Determine the canonical remote and PR target:
upstream exists, use it as the canonical remote. Fetch from it, base branches on it, and submit PRs to it.upstream remote exists, fall back to origin.Resolve the canonical repo's owner/name from the remote URL and store it. Use --repo <owner/name> on every gh command.
Run git fetch <canonical-remote> to ensure up-to-date refs.
If the issue references unfamiliar APIs, libraries, error messages, or domain concepts, search for official documentation and known solutions before planning. Skip for straightforward bugs where the fix is clear from the codebase alone.
Write a PRD covering:
Stop and wait for user approval. Do not proceed until the user affirms (e.g., "approved", "go", "lgtm").
Determine the branch prefix from issue type:
fix/ for bugsfeat/ for featuresrefactor/ for refactorsdocs/ for documentationfix/Create a branch named {prefix}issue-$ARGUMENTS based on <canonical-remote>/<default-branch>.
Implement the plan across all necessary files. Follow the project's CLAUDE.md standards. Keep changes minimal and focused on the issue requirements — no speculative features.
Tests are part of implementation, not a separate step. Add tests for the changed behavior alongside the code changes.
When stuck — a confusing error, an unfamiliar API, or an approach that isn't working — search for solutions rather than spinning.
CI is the source of truth. Before running anything, read the project's CI configuration to learn what the project actually runs.
.github/workflows/ for the main CI workflow. Extract:
git diff --exit-code)Makefile or justfile if present. Cross-reference targets used in CI.Store discovered commands. They override fallback defaults.
Detect the project language from manifest files. A project may use multiple languages; run checks for each.
Run in this order, using CI-discovered commands when available:
git diff --exit-code. Regenerate stale files if needed.For fallback commands when CI doesn't specify what to run, see references/quality-gate-defaults.md.
Iterate on failures until green.
Review your own diff against the base branch. Rank every finding:
Address all P1–P3 findings. For each finding, either:
After fixing, re-read the diff of changes made in this step and verify each fix is correct and doesn't introduce regressions.
Re-run the full quality pipeline from step 7b. Iterate until clean.
Invoke /vc-ship. The PR description should:
Closes #$ARGUMENTS (or Refs # if it doesn't fully close it)vc-shipobsidian-releasevc-synctesting
Audits ~/.claude/skills/ for unused entries, duplicate names, missing descriptions, and the longest descriptions. Use when trimming the user-level skill set, asking which skills are unused, finding duplicates, or auditing skill hygiene.
tools
Publishes and manages Flowershow sites with the `fl` CLI (the Go-based successor to the deprecated `@flowershow/publish` npm package). Use when publishing a note or folder to Flowershow, syncing updates to an existing site, managing auth, listing or deleting sites, or installing/upgrading the CLI.
tools
Copy edits prose while preserving voice and register. Use when asked to edit, copy edit, line edit, proofread, revise, polish, tighten, rewrite, or clean up essays, articles, drafts, or fiction. Flags wordiness, passive voice, clichés, hedging, and nominalizations with bracket markup or clean rewrites.
tools
Improves CLAUDE.md by analyzing conversation patterns. Use when Claude keeps repeating a mistake, when teaching a new preference, or when consolidating guidance from repeated instructions. Captures recurring corrections and style preferences into project instructions.