Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Resolve a fuzzy GitHub reference (issue number, PR number, or text description) to a URL and optionally open it in the browser.
.claude/skills/sonichi-open-sutando-ref/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-20 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 22% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -58% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -34% | 0% |
Resolves natural-language and numeric GitHub references to URLs and optionally navigates to them. Closes #903.
Usage: /open-sutando-ref [--open] <ref>
ARGUMENTS: $ARGUMENTS
#874 / 874 — numeric (auto-detects PR vs issue)PR 874 / pull request 874 — explicit PRissue 874 — explicit issue"the result-marker PR" — fuzzy text search via gh pr/issue list --search"multi-core scheduler" — searches both PRs and issues, returns best matchgh pr list --search <query> then gh issue list --search <query>; return the top match URL.--open flag is present, call open <url> to navigate in the default browser.SUTANDO_GH_REPO — override the target repo (default: inferred from gh repo view or sonichi/sutando).resolve.py --open "result-marker PR"resolve.py "#874"| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-20 | fail→pass | 9,250 | 2,169 | -77% | 1 | 1 | 0% | 1,425 | 742 | -48% | 0 | 0 | — |
case-01 | fail→pass | 5,894 | 10,947 | +86% | 1 | 1 | 0% | 1,072 | 1,043 | -3% | 0 | 0 | — |
case-02 | fail→pass | 4,346 | 2,555 | -41% | 1 | 1 | 0% | 624 | 761 | +22% | 0 | 0 | — |
case-03 | fail→fail | 3,617 | 5,862 | +62% | 1 | 1 | 0% | 550 | 773 | +41% | 0 | 0 | — |
case-04 | pass→pass | 9,159 | 11,220 | +23% | 1 | 1 | 0% | 1,378 | 1,051 | -24% | 0 | 0 | — |
case-05 | pass→fail | 5,443 | 5,753 | +6% | 1 | 1 | 0% | 873 | 523 | -40% | 0 | 0 | — |
case-06 | pass→pass | 4,478 | 7,133 | +59% | 1 | 1 | 0% | 736 | 1,616 | +120% | 0 | 0 | — |
case-11 | pass→fail | 7,084 | 17,118 | +142% | 1 | 1 | 0% | 1,134 | 875 | -23% | 0 | 0 | — |
case-07 | fail→fail | 4,879 | 6,097 | +25% | 1 | 1 | 0% | 837 | 642 | -23% | 0 | 0 | — |
case-08 | pass→pass | 3,027 | 2,728 | -10% | 1 | 1 | 0% | 446 | 859 | +93% | 0 | 0 | — |
case-09 | pass→pass | 3,101 | 3,382 | +9% | 1 | 1 | 0% | 445 | 893 | +101% | 0 | 0 | — |
case-10 | fail→fail | 6,010 | 5,014 | -17% | 1 | 1 | 0% | 858 | 629 | -27% | 0 | 0 | — |
case-12 | fail→fail | 3,995 | 9,385 | +135% | 1 | 1 | 0% | 576 | 844 | +47% | 0 | 0 | — |
case-13 | fail→pass | 10,459 | 2,623 | -75% | 1 | 1 | 0% | 1,977 | 825 | -58% | 0 | 0 | — |
case-14 | pass→pass | 2,867 | 2,039 | -29% | 1 | 1 | 0% | 473 | 762 | +61% | 0 | 0 | — |
case-15 | pass→pass | 2,977 | 2,578 | -13% | 1 | 1 | 0% | 460 | 795 | +73% | 0 | 0 | — |
case-16 | pass→pass | 7,010 | 2,719 | -61% | 1 | 1 | 0% | 1,099 | 837 | -24% | 0 | 0 | — |
case-17 | fail→pass | 7,867 | 2,798 | -64% | 1 | 1 | 0% | 1,271 | 835 | -34% | 0 | 0 | — |
case-18 | pass→pass | 10,473 | 5,422 | -48% | 1 | 1 | 0% | 1,692 | 1,407 | -17% | 0 | 0 | — |
case-19 | fail→pass | 6,243 | 1,715 | -73% | 1 | 1 | 0% | 1,032 | 647 | -37% | 0 | 0 | — |
case-21 | pass→pass | 11,077 | 2,208 | -80% | 1 | 1 | 0% | 1,753 | 705 | -60% | 0 | 0 | — |
case-22 | pass→pass | 5,705 | 1,430 | -75% | 1 | 1 | 0% | 858 | 603 | -30% | 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 16 counted toward the lift figure. The other 6 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 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.