Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Operate GitHub issues, pull requests, branches, checks, workflows, and permissions through Rube MCP. Use when GitHub work must be queried or changed programmatically with repository-policy safeguards.
.claude/skills/sickn33-github-automation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | 42% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-10 | ✓→✗ | ▼ Worse | 74% | 0% |
| case-12 | ✓→✗ | ▼ Worse | -36% | 0% |
| case-15 | ✓→✗ | ▼ Worse | -34% | 0% |
Use Composio's GitHub toolkit through Rube MCP while preserving repository policy, exact revision identity, and branch protection.
Use for programmatic GitHub issue, pull-request, branch, Actions, deployment, collaborator, or protection tasks when Rube MCP is available. Prefer the native gh workflow or a repository-specific maintainer command when local repository policy requires it.
RUBE_SEARCH_TOOLS is available.RUBE_MANAGE_CONNECTIONS with toolkit github and complete OAuth only if the connection is not active.owner/repo; do not rely on a similarly named repository.Never request, print, or persist GitHub credentials in prompts or artifacts.
Before mutation, read AGENTS.md, contribution and maintainer docs, then inspect the default branch and effective protection. Repository-native commands and required checks take precedence over generic Rube operations.
If a repository provides a guarded merge or release command, use it instead of the generic merge tool. In agentic-awesome-skills, use antigravity-maintainer-batch-release and npm run merge:batch so exact-SHA review, fresh check-suite binding, and protected main are enforced.
Paginate until the requested result set is complete. Treat silent omission of labels or assignees as a permissions failure, not success.
MERGED and confirm the target branch contains the intended commit.Do not treat a successful API call that enables auto-merge or queues work as an immediate merge.
Deletion, force-push, default-branch changes, and protection changes are destructive or high-impact actions requiring explicit authorization.
workflow_dispatch support before dispatch.queued, in_progress, action_required, completed, and skipped states.Do not approve fork workflow runs by raw run ID when the repository provides a guarded approval command.
User request:
> Operate GitHub issues, pull requests, branches, checks, workflows, and permissions through Rube MCP.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→fail | 6,707 | 4,922 | -27% | 1 | 1 | 0% | 506 | 1,263 | +150% | 0 | 0 | — |
case-01 | fail→fail | 7,750 | 3,838 | -50% | 1 | 1 | 0% | 413 | 1,506 | +265% | 0 | 0 | — |
case-02 | fail→fail | 6,201 | 5,996 | -3% | 1 | 1 | 0% | 201 | 1,420 | +606% | 0 | 0 | — |
case-04 | fail→fail | 5,851 | 10,177 | +74% | 1 | 1 | 0% | 395 | 2,756 | +598% | 0 | 0 | — |
case-05 | fail→fail | 16,079 | 8,251 | -49% | 1 | 1 | 0% | 1,602 | 1,545 | -4% | 0 | 0 | — |
case-06 | fail→fail | 15,222 | 5,077 | -67% | 1 | 1 | 0% | 2,230 | 1,255 | -44% | 0 | 0 | — |
case-07 | pass→pass | 6,751 | 4,164 | -38% | 1 | 1 | 0% | 871 | 1,631 | +87% | 0 | 0 | — |
case-08 | fail→fail | 12,302 | 9,215 | -25% | 1 | 1 | 0% | 2,094 | 1,205 | -42% | 0 | 0 | — |
case-09 | pass→pass | 6,436 | 8,881 | +38% | 1 | 1 | 0% | 1,248 | 2,505 | +101% | 0 | 0 | — |
case-10 | pass→fail | 5,854 | 3,453 | -41% | 1 | 1 | 0% | 853 | 1,486 | +74% | 0 | 0 | — |
case-11 | fail→pass | 14,303 | 13,680 | -4% | 1 | 1 | 0% | 1,636 | 2,324 | +42% | 0 | 0 | — |
case-12 | pass→fail | 12,144 | 6,030 | -50% | 1 | 1 | 0% | 2,182 | 1,400 | -36% | 0 | 0 | — |
case-13 | pass→pass | 9,213 | 7,569 | -18% | 1 | 1 | 0% | 1,615 | 2,238 | +39% | 0 | 0 | — |
case-14 | pass→pass | 8,463 | 5,783 | -32% | 1 | 1 | 0% | 1,475 | 1,825 | +24% | 0 | 0 | — |
case-15 | pass→fail | 12,792 | 4,628 | -64% | 1 | 1 | 0% | 2,005 | 1,329 | -34% | 0 | 0 | — |
case-16 | fail→pass | 11,397 | 13,181 | +16% | 1 | 1 | 0% | 1,815 | 2,657 | +46% | 0 | 0 | — |
case-17 | pass→pass | 14,714 | 8,453 | -43% | 1 | 1 | 0% | 2,283 | 2,366 | +4% | 0 | 0 | — |
case-18 | pass→fail | 3,106 | 5,866 | +89% | 1 | 1 | 0% | 410 | 1,312 | +220% | 0 | 0 | — |
case-19 | fail→fail | 6,456 | 7,663 | +19% | 1 | 1 | 0% | 969 | 2,340 | +141% | 0 | 0 | — |
case-20 | pass→pass | 11,760 | 7,693 | -35% | 1 | 1 | 0% | 2,085 | 2,329 | +12% | 0 | 0 | — |
case-21 | pass→pass | 4,054 | 5,057 | +25% | 1 | 1 | 0% | 777 | 1,937 | +149% | 0 | 0 | — |
case-22 | pass→pass | 6,671 | 4,696 | -30% | 1 | 1 | 0% | 1,102 | 1,713 | +55% | 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 13 counted toward the lift figure. The other 9 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 -9 percentage points is the difference between those two pass rates over the 13 comparable cases. 4 cases got worse with the skill loaded, and they are included in that figure.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
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.