Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use the connected iPolloWork GitHub service to inspect repositories, pull requests, issues, reviews, and Actions, then route focused work to the review, CI, or publish skill.
.claude/skills/devin-axis-github/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -78% | 0% |
Use this skill as the entrypoint for general GitHub work. It combines structured remote GitHub data from the iPolloWork GitHub service with the current local checkout.
ipollowork_extension_list_actions with extensionId: "github" before using the service.ipollowork_extension_call only with the declared GitHub actions and explicit owner and repo values.git for branch, status, diff, commit, and push operations.github-review-follow-up.github-ci-debug.github-publish-changes.repository-context to confirm the target and default branch.list-pull-requests, pull-request-detail, list-issues, or issue-detail for the requested scope.Do not invent repository search results. If the repository cannot be derived from the request or current checkout, ask for owner/repo.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,256 | 16,555 | +128% | 1 | 1 | 0% | 302 | 738 | +144% | 0 | 0 | — |
case-02 | fail→fail | 16,030 | 14,378 | -10% | 1 | 1 | 0% | 1,585 | 668 | -58% | 0 | 0 | — |
case-03 | fail→fail | 16,228 | 16,732 | +3% | 1 | 1 | 0% | 338 | 942 | +179% | 0 | 0 | — |
case-04 | fail→fail | 8,603 | 16,140 | +88% | 1 | 1 | 0% | 385 | 885 | +130% | 0 | 0 | — |
case-05 | fail→fail | 8,070 | 16,913 | +110% | 1 | 1 | 0% | 437 | 800 | +83% | 0 | 0 | — |
case-06 | fail→fail | 10,804 | 16,576 | +53% | 1 | 1 | 0% | 902 | 739 | -18% | 0 | 0 | — |
case-07 | pass→pass | 14,440 | 8,645 | -40% | 1 | 1 | 0% | 1,486 | 940 | -37% | 0 | 0 | — |
case-08 | fail→pass | 12,059 | 10,476 | -13% | 1 | 1 | 0% | 1,250 | 956 | -24% | 0 | 0 | — |
case-09 | pass→pass | 11,324 | 8,606 | -24% | 1 | 1 | 0% | 1,016 | 1,022 | +1% | 0 | 0 | — |
case-10 | fail→pass | 9,566 | 7,275 | -24% | 1 | 1 | 0% | 685 | 726 | +6% | 0 | 0 | — |
case-11 | fail→fail | 10,004 | 8,128 | -19% | 1 | 1 | 0% | 826 | 901 | +9% | 0 | 0 | — |
case-12 | fail→pass | 13,924 | 7,922 | -43% | 1 | 1 | 0% | 1,546 | 878 | -43% | 0 | 0 | — |
case-13 | fail→pass | 26,218 | 8,017 | -69% | 1 | 1 | 0% | 1,722 | 926 | -46% | 0 | 0 | — |
case-14 | fail→pass | 25,067 | 15,261 | -39% | 1 | 1 | 0% | 3,566 | 780 | -78% | 0 | 0 | — |
case-15 | fail→pass | 15,489 | 15,835 | +2% | 1 | 1 | 0% | 1,629 | 907 | -44% | 0 | 0 | — |
case-16 | fail→fail | 12,258 | 12,795 | +4% | 1 | 1 | 0% | 1,317 | 1,647 | +25% | 0 | 0 | — |
case-17 | fail→fail | 18,684 | 6,612 | -65% | 1 | 1 | 0% | 2,218 | 634 | -71% | 0 | 0 | — |
case-18 | pass→pass | 2,976 | 8,199 | +176% | 1 | 1 | 0% | 443 | 873 | +97% | 0 | 0 | — |
case-19 | pass→pass | 13,923 | 8,006 | -42% | 1 | 1 | 0% | 1,360 | 856 | -37% | 0 | 0 | — |
case-20 | pass→pass | 9,564 | 7,637 | -20% | 1 | 1 | 0% | 717 | 839 | +17% | 0 | 0 | — |
case-21 | fail→fail | 12,264 | 7,256 | -41% | 1 | 1 | 0% | 1,058 | 749 | -29% | 0 | 0 | — |
case-22 | pass→fail | 10,382 | 14,464 | +39% | 1 | 1 | 0% | 800 | 560 | -30% | 0 | 0 | — |
case-23 | fail→pass | 14,431 | 8,109 | -44% | 1 | 1 | 0% | 1,654 | 929 | -44% | 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. 23 cases were attempted, and 16 counted toward the lift figure. The other 7 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 +26 percentage points is the difference between those two pass rates over the 16 comparable cases. 2 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.