Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Step through versioning, tagging, and verification
.claude/skills/devin-axis-release/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -70% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -51% | 0% |
Confirm the repo is on main and clean. Keep changes aligned with OpenCode primitives like .opencode, opencode.json, skills, and plugins when relevant.
Update versions in packages/app/package.json, packages/desktop/package.json, packages/orchestrator/package.json (publishes as ipollowork-orchestrator), packages/desktop/src-tauri/tauri.conf.json, and packages/desktop/src-tauri/Cargo.toml. Use one of these commands.
bashpnpm bump:patch pnpm bump:minor pnpm bump:major pnpm bump:set -- 0.1.21
Merge the version bump into main. Make sure no secrets or credentials are committed.
Create and push the tag to trigger the Release App workflow.
bashgit tag vX.Y.Z git push origin vX.Y.Z
If a tag needs a rerun, dispatch the workflow.
bashgh workflow run "Release App" --repo Devin-AXIS/iPolloWork -f tag=vX.Y.Z
Confirm the run and the published release.
bashgh run list --repo Devin-AXIS/iPolloWork --workflow "Release App" --limit 5 gh release view vX.Y.Z --repo Devin-AXIS/iPolloWork
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 14,740 | 15,108 | +2% | 1 | 1 | 0% | 1,709 | 616 | -64% | 0 | 0 | — |
case-02 | fail→fail | 16,181 | 13,721 | -15% | 1 | 1 | 0% | 358 | 557 | +56% | 0 | 0 | — |
case-03 | fail→fail | 11,534 | 7,977 | -31% | 1 | 1 | 0% | 1,134 | 681 | -40% | 0 | 0 | — |
case-04 | pass→fail | 17,914 | 14,577 | -19% | 1 | 1 | 0% | 2,195 | 613 | -72% | 0 | 0 | — |
case-05 | pass→pass | 18,160 | 13,887 | -24% | 1 | 1 | 0% | 2,367 | 1,915 | -19% | 0 | 0 | — |
case-06 | pass→pass | 17,153 | 14,092 | -18% | 1 | 1 | 0% | 2,292 | 2,023 | -12% | 0 | 0 | — |
case-07 | fail→pass | 14,191 | 7,655 | -46% | 1 | 1 | 0% | 1,706 | 769 | -55% | 0 | 0 | — |
case-08 | fail→pass | 17,828 | 6,813 | -62% | 1 | 1 | 0% | 2,361 | 710 | -70% | 0 | 0 | — |
case-09 | fail→pass | 18,138 | 9,802 | -46% | 1 | 1 | 0% | 2,299 | 1,282 | -44% | 0 | 0 | — |
case-10 | fail→pass | 12,677 | 7,620 | -40% | 1 | 1 | 0% | 1,441 | 778 | -46% | 0 | 0 | — |
case-11 | fail→pass | 11,926 | 6,805 | -43% | 1 | 1 | 0% | 1,300 | 639 | -51% | 0 | 0 | — |
case-12 | fail→pass | 14,068 | 7,134 | -49% | 1 | 1 | 0% | 1,665 | 775 | -53% | 0 | 0 | — |
case-13 | fail→pass | 18,946 | 6,932 | -63% | 1 | 1 | 0% | 2,387 | 669 | -72% | 0 | 0 | — |
case-14 | fail→pass | 12,564 | 7,061 | -44% | 1 | 1 | 0% | 1,230 | 665 | -46% | 0 | 0 | — |
case-15 | pass→pass | 9,645 | 7,331 | -24% | 1 | 1 | 0% | 680 | 744 | +9% | 0 | 0 | — |
case-16 | fail→pass | 19,207 | 6,642 | -65% | 1 | 1 | 0% | 2,313 | 586 | -75% | 0 | 0 | — |
case-17 | fail→pass | 13,043 | 7,094 | -46% | 1 | 1 | 0% | 1,350 | 660 | -51% | 0 | 0 | — |
case-22 | pass→fail | 10,449 | 7,591 | -27% | 1 | 1 | 0% | 887 | 740 | -17% | 0 | 0 | — |
case-18 | pass→pass | 12,982 | 6,711 | -48% | 1 | 1 | 0% | 1,278 | 607 | -53% | 0 | 0 | — |
case-19 | pass→pass | 12,893 | 7,506 | -42% | 1 | 1 | 0% | 1,142 | 553 | -52% | 0 | 0 | — |
case-20 | fail→fail | 10,026 | 7,269 | -27% | 1 | 1 | 0% | 913 | 786 | -14% | 0 | 0 | — |
case-21 | pass→pass | 11,321 | 6,690 | -41% | 1 | 1 | 0% | 1,086 | 627 | -42% | 0 | 0 | — |
case-23 | fail→pass | 42,855 | 7,176 | -83% | 1 | 1 | 0% | 7,171 | 678 | -91% | 0 | 0 | — |
case-24 | fail→pass | 8,817 | 6,847 | -22% | 1 | 1 | 0% | 670 | 618 | -8% | 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. 24 cases were attempted, and 21 counted toward the lift figure. The other 3 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 +42 percentage points is the difference between those two pass rates over the 21 comparable cases. 3 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.