Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Encourages proactive use of GitHub CLI (gh) for gathering context on PRs, issues, comments, and repository information when working with GitHub-related tasks.
.claude/skills/aiskillstore-github-cli/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 322% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 56% | 0% |
| case-05 | ✓→✓ | = Same ✓ | 32% | 0% |
| case-06 | ✓→✓ | = Same ✓ | 148% | 0% |
| case-07 | ✓→✓ | = Same ✓ | 94% | 0% |
This skill encourages proactive use of the GitHub CLI (gh) to gather rich context when working with GitHub-related tasks.
When the user mentions PRs, issues, branches, code reviews, comments, or anything GitHub-related, proactively use gh commands to gather context rather than relying solely on local git commands.
Local git tells you about commits and branches. GitHub CLI tells you about the conversation around those changes — PR descriptions, review comments, issue discussions, CI status, and more.
Use gh commands when the user mentions or asks about:
github.com URL can be inspected via ghbash# View PR details (description, status, checks) gh pr view PR_NUMBER # View PR diff gh pr diff PR_NUMBER # List PR comments gh api repos/OWNER/REPO/pulls/PR_NUMBER/comments # List review comments (inline code comments) gh api repos/OWNER/REPO/pulls/PR_NUMBER/reviews # Check PR status and CI checks gh pr checks PR_NUMBER # List open PRs gh pr list # List PRs by author gh pr list --author USERNAME
bash# View issue details gh issue view ISSUE_NUMBER # List issue comments gh api repos/OWNER/REPO/issues/ISSUE_NUMBER/comments # List open issues gh issue list # Search issues gh issue list --search "QUERY"
bash# View repo details gh repo view # List branches gh api repos/OWNER/REPO/branches # View recent releases gh release list # View workflow runs gh run list
When given a GitHub URL, extract the relevant information and use gh:
bash# From: https://github.com/monzo/analytics/pull/123 gh pr view 123 --repo monzo/analytics # From: https://github.com/monzo/analytics/issues/456 gh issue view 456 --repo monzo/analytics
bash# Get the full picture gh pr view PR_NUMBER # Description and status gh pr diff PR_NUMBER # What changed gh pr checks PR_NUMBER # CI status gh api repos/OWNER/REPO/pulls/PR_NUMBER/comments # Discussion
bashgh issue view ISSUE_NUMBER # Issue details gh api repos/OWNER/REPO/issues/ISSUE_NUMBER/comments # Discussion
bash# What PRs exist for this branch? gh pr list --head BRANCH_NAME # What's the status of my PR? gh pr status
Combine gh with local git for full context:
bash# Local: What commits are on this branch? git log origin/master..HEAD --oneline # GitHub: What's the PR discussion saying? gh pr view --comments
--json for structured output when you need to parse data:bash gh pr view PR_NUMBER --json title,body,reviews,comments
gh api for anything not covered by high-level commands — it gives direct access to the GitHub API--repo OWNER/REPO when working outside the repo directory or when ambiguousThis skill should guide behaviour whenever GitHub-related context would be valuable. You don't need to explicitly invoke it — just remember to reach for gh when the user mentions:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 3,150 | 14,469 | +359% | 1 | 1 | 0% | 421 | 1,778 | +322% | 0 | 0 | — |
case-02 | fail→fail | 7,033 | 16,506 | +135% | 1 | 1 | 0% | 697 | 1,255 | +80% | 0 | 0 | — |
case-03 | fail→fail | 12,378 | 16,223 | +31% | 1 | 1 | 0% | 526 | 1,303 | +148% | 0 | 0 | — |
case-04 | pass→pass | 12,566 | 8,221 | -35% | 1 | 1 | 0% | 977 | 1,528 | +56% | 0 | 0 | — |
case-05 | pass→pass | 16,162 | 5,656 | -65% | 1 | 1 | 0% | 1,577 | 2,082 | +32% | 0 | 0 | — |
case-06 | pass→pass | 10,599 | 7,626 | -28% | 1 | 1 | 0% | 568 | 1,408 | +148% | 0 | 0 | — |
case-07 | pass→pass | 10,162 | 6,900 | -32% | 1 | 1 | 0% | 665 | 1,293 | +94% | 0 | 0 | — |
case-08 | pass→pass | 14,650 | 11,203 | -24% | 1 | 1 | 0% | 1,297 | 1,727 | +33% | 0 | 0 | — |
case-09 | pass→pass | 12,035 | 8,210 | -32% | 1 | 1 | 0% | 1,269 | 1,626 | +28% | 0 | 0 | — |
case-10 | pass→pass | 12,853 | 18,039 | +40% | 1 | 1 | 0% | 1,238 | 1,575 | +27% | 0 | 0 | — |
case-11 | pass→pass | 9,970 | 9,172 | -8% | 1 | 1 | 0% | 871 | 1,542 | +77% | 0 | 0 | — |
case-12 | pass→pass | 5,826 | 18,153 | +212% | 1 | 1 | 0% | 1,000 | 1,681 | +68% | 0 | 0 | — |
case-13 | pass→pass | 24,718 | 12,395 | -50% | 1 | 1 | 0% | 1,237 | 1,696 | +37% | 0 | 0 | — |
case-14 | pass→pass | 12,686 | 10,089 | -20% | 1 | 1 | 0% | 1,111 | 1,663 | +50% | 0 | 0 | — |
case-15 | pass→pass | 12,340 | 6,424 | -48% | 1 | 1 | 0% | 1,204 | 1,938 | +61% | 0 | 0 | — |
case-16 | pass→pass | 15,603 | 11,379 | -27% | 1 | 1 | 0% | 1,876 | 2,129 | +13% | 0 | 0 | — |
case-17 | pass→pass | 23,723 | 8,218 | -65% | 1 | 1 | 0% | 355 | 1,348 | +280% | 0 | 0 | — |
case-18 | pass→pass | 11,591 | 8,231 | -29% | 1 | 1 | 0% | 962 | 1,426 | +48% | 0 | 0 | — |
case-19 | pass→pass | 5,599 | 9,626 | +72% | 1 | 1 | 0% | 715 | 1,551 | +117% | 0 | 0 | — |
case-20 | pass→pass | 11,099 | 10,230 | -8% | 1 | 1 | 0% | 1,100 | 1,972 | +79% | 0 | 0 | — |
case-21 | pass→pass | 15,227 | 12,919 | -15% | 1 | 1 | 0% | 1,462 | 2,051 | +40% | 0 | 0 | — |
case-22 | pass→pass | 26,618 | 9,684 | -64% | 1 | 1 | 0% | 1,354 | 1,792 | +32% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +5 percentage points is the difference between those two pass rates over the 20 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.