Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate daily standup notes from git activity
.claude/skills/hashgraph-online-standup-gen/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 34 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -59% | 0% |
Role: act as a standup-note author. Produce Yesterday / Today / Blockers sections grounded only in the user's git activity.
Run the helper:
bashpython3 scripts/standup.py --format md
Useful flags: --since 2026-05-15, --since yesterday, --repos /path/repoA,/path/repoB, --author all, --format json.
The helper fills "Yesterday" from commit subjects and leaves "Today (planned)" and "Blockers" as placeholders. Read the commit list, then propose 1–3 candidate "Today" items derived from in-flight files (e.g. files with WIP or follow-up commits). Mark blockers only if the diffs or commit messages clearly evidence one. Print the final markdown.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 14,110 | 13,906 | -1% | 1 | 1 | 0% | 1,500 | 276 | -82% | 0 | 0 | — |
case-02 | fail→fail | 11,511 | 14,251 | +24% | 1 | 1 | 0% | 1,101 | 385 | -65% | 0 | 0 | — |
case-03 | fail→fail | 9,968 | 4,148 | -58% | 1 | 1 | 0% | 850 | 390 | -54% | 0 | 0 | — |
case-04 | fail→fail | 12,361 | 11,089 | -10% | 1 | 1 | 0% | 2,284 | 632 | -72% | 0 | 0 | — |
case-05 | fail→pass | 13,496 | 3,632 | -73% | 1 | 1 | 0% | 1,466 | 791 | -46% | 0 | 0 | — |
case-06 | fail→pass | 14,814 | 10,103 | -32% | 1 | 1 | 0% | 1,786 | 1,102 | -38% | 0 | 0 | — |
case-07 | fail→fail | 13,835 | 7,318 | -47% | 1 | 1 | 0% | 1,511 | 582 | -61% | 0 | 0 | — |
case-08 | fail→fail | 16,554 | 14,460 | -13% | 1 | 1 | 0% | 2,136 | 658 | -69% | 0 | 0 | — |
case-09 | pass→fail | 11,527 | 4,533 | -61% | 1 | 1 | 0% | 1,730 | 929 | -46% | 0 | 0 | — |
case-10 | fail→fail | 14,235 | 16,005 | +12% | 1 | 1 | 0% | 1,425 | 677 | -52% | 0 | 0 | — |
case-11 | pass→pass | 8,189 | 3,896 | -52% | 1 | 1 | 0% | 1,277 | 863 | -32% | 0 | 0 | — |
case-12 | fail→pass | 7,468 | 10,681 | +43% | 1 | 1 | 0% | 1,285 | 1,188 | -8% | 0 | 0 | — |
case-13 | fail→fail | 7,065 | 9,289 | +31% | 1 | 1 | 0% | 1,119 | 952 | -15% | 0 | 0 | — |
case-14 | pass→pass | 17,771 | 9,975 | -44% | 1 | 1 | 0% | 2,120 | 1,111 | -48% | 0 | 0 | — |
case-15 | fail→fail | 8,862 | 7,946 | -10% | 1 | 1 | 0% | 1,415 | 716 | -49% | 0 | 0 | — |
case-16 | fail→pass | 12,394 | 3,191 | -74% | 1 | 1 | 0% | 1,190 | 727 | -39% | 0 | 0 | — |
case-17 | fail→pass | 14,643 | 3,016 | -79% | 1 | 1 | 0% | 1,629 | 673 | -59% | 0 | 0 | — |
case-18 | fail→fail | 11,586 | 12,275 | +6% | 1 | 1 | 0% | 1,887 | 368 | -80% | 0 | 0 | — |
case-19 | pass→pass | 10,727 | 18,143 | +69% | 1 | 1 | 0% | 1,035 | 579 | -44% | 0 | 0 | — |
case-20 | pass→pass | 15,581 | 6,861 | -56% | 1 | 1 | 0% | 1,465 | 436 | -70% | 0 | 0 | — |
case-21 | pass→pass | 14,174 | 15,821 | +12% | 1 | 1 | 0% | 2,394 | 2,910 | +22% | 0 | 0 | — |
case-22 | pass→pass | 16,319 | 14,352 | -12% | 1 | 1 | 0% | 2,217 | 1,866 | -16% | 0 | 0 | — |
case-23 | pass→fail | 20,672 | 25,499 | +23% | 1 | 1 | 0% | 2,676 | 4,668 | +74% | 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 +13 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.