kocal/dotfiles/github-cli
- community
Checking that the SKILL.md file is still in the repository.
Fetch real GitHub data with the `gh` CLI instead of reconstructing it from memory. Use whenever a message mentions an issue or PR number (`#123`, "issue 789", "la PR 12"), pastes a github.com URL, or brings up CI failures, review comments, checks, labels, releases or tags, even when the word "GitHub" never appears, since "the CI is red" and a bare `#456` are already GitHub references. Use it too when opening a pull request (repo template, fork workflow, `gh pr create` traps) or checking out someone else's PR to rebase it or answer its review.
Install
npx vskill@latest install kocal/dotfiles/github-cliSKILL.md
GitHub CLI
gh is the only reliable account of what an issue, a PR, a review or a CI run actually contains. Reconstructing that from memory, from a branch name, or from what the user seems to imply produces fiction that reads exactly like fact, and they usually find out only after acting on it. Fetch first, answer second.
Rules
Run `gh` before answering, even when the question looks answerable from context. "Is #42 still open?" still needs a fetch: the state moved since anything you might remember, and a stale "yes" costs more than a slow one.
Only emit URLs that came from the user or from `gh` output. A github.com/owner/repo/pull/123 you assembled yourself is a 404 at best and a stranger's PR at worst.
Reach for the specific subcommand before `gh api`. gh issue view, gh pr checks and gh run view return output shaped for reading. gh api returns raw JSON where the one field you wanted sits among fifty you did not, all of it billed to your context. Fall back to gh api when no subcommand covers the case, and pair it with --jq when you do.
Quote the part you acted on: the failing log line, the review comment, the check name. The user should be able to audit your conclusion without re-running your commands.
Workflow rules live in `CLAUDE.md`, not here. Its Git & PR and Merging is mine, never yours sections govern fork versus upstream, one PR = one commit, the PR template, and the fact that pushing and merging belong to the user alone. This file covers gh mechanics only. Where the two look like they disagree, CLAUDE.md wins.
Reading issues and pull requests
`--comments` stops at the conversation tab. Comments left on specific diff lines are a separate resource, and they are usually where the actual review findings are. On one symfony/symfony PR, gh pr view --comments surfaced 6 entries where the inline endpoint returned 28. Answering a review off the first number means missing most of it:
line is null once the diff hunk a comment was attached to has moved; original_line still points at the code it was written against.
Reading CI
`--log` dumps the entire run, which on a matrix build is tens of thousands of lines landing in context in full. Start with --log-failed, and filter when you only need the assertion that broke:
Going further
| Task | Read | |
|---|---|---|
Open a PR: repo template, --fill, pushing from a fork | references/creating-prs.md | |
| Take over someone else's PR: check out, rebase, answer the review | references/reviewing-prs.md |
Anything with no subcommand goes through gh api, which speaks the whole REST API and, with graphql, the parts REST does not reach:
Worked examples
"What's the status of #42?" -> gh issue view 42, then report state, assignees and the last thing that happened on it. "It's open" is the part they already knew.
"The CI is red on my PR." -> gh pr checks to find which check failed and its run id, then gh run view <id> --log-failed for the reason. Quote the failing assertion and say what would fix it.
A bare `https://github.com/acme/app/pull/789` with no question attached -> gh pr view 789 --repo acme/app. If they asked nothing, summarise state, checks, and what the PR is waiting on.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingnix/home/claude/skills/github-cli/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/kocal/dotfiles/github-cli)