Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Inspect GitHub Actions self-hosted runner queues when a developer provides a run URL, job URL, runner label, or reports a blocked CI job, or invokes /qv-devops-runner-queue.
.claude/skills/tetherto-qv-devops-runner-queue/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 324% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-05 | ✓→✗ | ▼ Worse | 3% | 0% |
| case-06 | ✓→✗ | ▼ Worse | -23% | 0% |
Run the read-only helper once and present its Markdown output directly without reformatting or a code fence:
bashnode .cursor/skills/qv-devops-runner-queue/scripts/inspect-runner-queue.mjs \ "<job-url|run-url|runner-label>" \ --repo "owner/repo"
For labels, --repo defaults to tetherto/qvac; URLs supply their repository. Add --json only when requested. Prefer a job URL for its complete scheduling labels. A run URL must contain exactly one active self-hosted job; otherwise show the helper's candidates and ask for a job URL.
If gh is missing or unauthenticated, report the helper error; never start authentication. Do not repeat a query unless the user requests a refresh.
The helper only makes GitHub GET requests. Never cancel, rerun, change runner state, or mutate a repository. Cancellation affects the whole run and requires a separate inspection and confirmation.
Interpretation:
created_at;--json keeps all collected jobs.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 6,326 | 31,047 | +391% | 1 | 1 | 0% | 1,102 | 4,674 | +324% | 0 | 0 | — |
case-02 | fail→fail | 6,399 | 4,946 | -23% | 1 | 1 | 0% | 1,047 | 629 | -40% | 0 | 0 | — |
case-03 | fail→fail | 7,869 | 5,517 | -30% | 1 | 1 | 0% | 515 | 693 | +35% | 0 | 0 | — |
case-04 | fail→fail | 4,791 | 7,021 | +47% | 1 | 1 | 0% | 887 | 891 | +0% | 0 | 0 | — |
case-05 | pass→fail | 3,970 | 7,224 | +82% | 1 | 1 | 0% | 768 | 792 | +3% | 0 | 0 | — |
case-06 | pass→fail | 5,267 | 7,330 | +39% | 1 | 1 | 0% | 958 | 742 | -23% | 0 | 0 | — |
case-07 | fail→fail | 7,985 | 11,842 | +48% | 1 | 1 | 0% | 1,346 | 528 | -61% | 0 | 0 | — |
case-08 | pass→fail | 5,019 | 5,376 | +7% | 1 | 1 | 0% | 924 | 584 | -37% | 0 | 0 | — |
case-09 | pass→fail | 8,426 | 5,445 | -35% | 1 | 1 | 0% | 1,503 | 691 | -54% | 0 | 0 | — |
case-10 | pass→fail | 5,046 | 5,316 | +5% | 1 | 1 | 0% | 866 | 676 | -22% | 0 | 0 | — |
case-11 | fail→fail | 9,496 | 5,343 | -44% | 1 | 1 | 0% | 1,683 | 701 | -58% | 0 | 0 | — |
case-12 | fail→fail | 9,796 | 5,298 | -46% | 1 | 1 | 0% | 1,951 | 663 | -66% | 0 | 0 | — |
case-13 | pass→pass | 2,973 | 9,328 | +214% | 1 | 1 | 0% | 486 | 1,918 | +295% | 0 | 0 | — |
case-14 | fail→fail | 8,051 | 4,745 | -41% | 1 | 1 | 0% | 1,392 | 694 | -50% | 0 | 0 | — |
case-15 | pass→fail | 6,495 | 4,509 | -31% | 1 | 1 | 0% | 1,207 | 591 | -51% | 0 | 0 | — |
case-16 | pass→pass | 11,059 | 2,359 | -79% | 1 | 1 | 0% | 1,809 | 658 | -64% | 0 | 0 | — |
case-17 | pass→pass | 8,188 | 1,767 | -78% | 1 | 1 | 0% | 1,308 | 579 | -56% | 0 | 0 | — |
case-18 | fail→pass | 9,259 | 1,239 | -87% | 1 | 1 | 0% | 1,293 | 486 | -62% | 0 | 0 | — |
case-19 | fail→pass | 9,036 | 1,711 | -81% | 1 | 1 | 0% | 1,394 | 527 | -62% | 0 | 0 | — |
case-20 | pass→pass | 8,571 | 1,463 | -83% | 1 | 1 | 0% | 1,464 | 532 | -64% | 0 | 0 | — |
case-21 | pass→pass | 11,530 | 1,640 | -86% | 1 | 1 | 0% | 1,746 | 526 | -70% | 0 | 0 | — |
case-22 | pass→pass | 12,263 | 4,039 | -67% | 1 | 1 | 0% | 2,035 | 902 | -56% | 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 9 counted toward the lift figure. The other 13 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 -14 percentage points is the difference between those two pass rates over the 9 comparable cases. 8 cases got worse with the skill loaded, and they are included in that figure.
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.