Install any skill in seconds. Free to start, no credit card required.
Get Started Free →GitHub Actions command center -- view workflow runs, read logs, re-run failed jobs, manage workflows, and debug CI failures entirely from the editor. Bypasses the deeply nested, visually-dependent Actions UI that is largely inaccessible to screen readers.
.claude/skills/community-access-actions-manager/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 142% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-08 | ✓→✓ | = Same ✓ | 35% | 0% |
Derived from .claude/agents/actions-manager.md. Treat platform-specific tool names or delegation instructions as Codex equivalents.
Shared instructions
Skills: github-workflow-standards, github-scanning
You are the Actions Manager. You give screen reader users and keyboard-only users full control over GitHub Actions workflows — a feature whose web UI presents deeply nested collapsible log trees, visual-only job dependency graphs, and conditionally-appearing controls that are largely inaccessible to assistive technology.
GitHub Actions UI presents severe accessibility barriers:
<pre> blocks that read as one giant unstructured text nodeThis agent bypasses all of that by working directly through the GitHub REST API.
gh api user.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,340 | 5,368 | -15% | 1 | 1 | 0% | 418 | 1,067 | +155% | 0 | 0 | — |
case-02 | fail→fail | 3,897 | 6,101 | +57% | 1 | 1 | 0% | 603 | 1,080 | +79% | 0 | 0 | — |
case-03 | fail→fail | 4,191 | 5,839 | +39% | 1 | 1 | 0% | 658 | 1,103 | +68% | 0 | 0 | — |
case-04 | fail→pass | 10,274 | 6,587 | -36% | 1 | 1 | 0% | 1,889 | 2,088 | +11% | 0 | 0 | — |
case-05 | fail→fail | 11,530 | 6,615 | -43% | 1 | 1 | 0% | 1,924 | 1,975 | +3% | 0 | 0 | — |
case-06 | fail→fail | 3,861 | 3,888 | +1% | 1 | 1 | 0% | 680 | 1,373 | +102% | 0 | 0 | — |
case-07 | fail→pass | 5,118 | 6,641 | +30% | 1 | 1 | 0% | 761 | 1,841 | +142% | 0 | 0 | — |
case-08 | pass→pass | 7,992 | 6,715 | -16% | 1 | 1 | 0% | 1,445 | 1,947 | +35% | 0 | 0 | — |
case-09 | fail→fail | 5,791 | 3,370 | -42% | 1 | 1 | 0% | 927 | 1,123 | +21% | 0 | 0 | — |
case-10 | fail→pass | 9,439 | 5,659 | -40% | 1 | 1 | 0% | 1,678 | 1,692 | +1% | 0 | 0 | — |
case-11 | fail→fail | 6,542 | 5,817 | -11% | 1 | 1 | 0% | 1,074 | 988 | -8% | 0 | 0 | — |
case-12 | fail→fail | 4,600 | 6,131 | +33% | 1 | 1 | 0% | 664 | 1,127 | +70% | 0 | 0 | — |
case-13 | fail→fail | 7,435 | 6,617 | -11% | 1 | 1 | 0% | 1,153 | 1,149 | -0% | 0 | 0 | — |
case-14 | fail→fail | 8,098 | 4,384 | -46% | 1 | 1 | 0% | 1,266 | 1,458 | +15% | 0 | 0 | — |
case-15 | fail→fail | 2,880 | 5,332 | +85% | 1 | 1 | 0% | 367 | 990 | +170% | 0 | 0 | — |
case-16 | fail→fail | 3,572 | 4,764 | +33% | 1 | 1 | 0% | 388 | 995 | +156% | 0 | 0 | — |
case-17 | pass→pass | 4,921 | 3,646 | -26% | 1 | 1 | 0% | 866 | 1,370 | +58% | 0 | 0 | — |
case-18 | fail→fail | 4,877 | 5,833 | +20% | 1 | 1 | 0% | 784 | 1,021 | +30% | 0 | 0 | — |
case-19 | fail→fail | 5,482 | 5,740 | +5% | 1 | 1 | 0% | 767 | 1,030 | +34% | 0 | 0 | — |
case-20 | fail→fail | 1,893 | 4,622 | +144% | 1 | 1 | 0% | 332 | 1,534 | +362% | 0 | 0 | — |
case-21 | fail→pass | 9,376 | 2,676 | -71% | 1 | 1 | 0% | 1,825 | 1,183 | -35% | 0 | 0 | — |
case-22 | pass→pass | 5,205 | 16,929 | +225% | 1 | 1 | 0% | 828 | 2,560 | +209% | 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 12 counted toward the lift figure. The other 10 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 +18 percentage points is the difference between those two pass rates over the 12 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.