Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →List TODO/FIXME/HACK comments with author + age in days
.claude/skills/hashgraph-online-todo-harvest/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -73% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -85% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -70% | 0% |
| case-22 | ✗→✓ | ▲ Improved | -37% | 0% |
Role: act as a debt-triage assistant. Surface the oldest, highest-cost TODOs first.
Run the helper:
bashpython3 scripts/harvest.py --format md
Useful flags: --markers TODO,FIXME,HACK, --min-age 90, --format json.
The helper uses git ls-files (respecting .gitignore) and runs git blame per match for author + age. Read the table, then propose a short triage list: which TODOs are stale enough to delete, which need owners, which look like real bugs. Be specific — quote the file and line.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→pass | 15,600 | 13,878 | -11% | 1 | 1 | 0% | 3,012 | 2,728 | -9% | 0 | 0 | — |
case-19 | pass→pass | 19,273 | 11,621 | -40% | 1 | 1 | 0% | 2,396 | 1,767 | -26% | 0 | 0 | — |
case-01 | fail→fail | 13,987 | 9,607 | -31% | 1 | 1 | 0% | 211 | 409 | +94% | 0 | 0 | — |
case-02 | fail→fail | 13,484 | 14,261 | +6% | 1 | 1 | 0% | 1,346 | 271 | -80% | 0 | 0 | — |
case-03 | fail→fail | 13,937 | 21,616 | +55% | 1 | 1 | 0% | 233 | 300 | +29% | 0 | 0 | — |
case-05 | fail→fail | 12,235 | 42,678 | +249% | 1 | 1 | 0% | 1,415 | 8,054 | +469% | 0 | 0 | — |
case-06 | pass→fail | 7,692 | 10,368 | +35% | 1 | 1 | 0% | 1,527 | 517 | -66% | 0 | 0 | — |
case-07 | fail→fail | 17,481 | 5,236 | -70% | 1 | 1 | 0% | 3,400 | 485 | -86% | 0 | 0 | — |
case-08 | pass→pass | 9,110 | 2,096 | -77% | 1 | 1 | 0% | 1,516 | 504 | -67% | 0 | 0 | — |
case-09 | fail→pass | 16,842 | 8,469 | -50% | 1 | 1 | 0% | 2,819 | 749 | -73% | 0 | 0 | — |
case-10 | pass→fail | 13,059 | 11,341 | -13% | 1 | 1 | 0% | 1,425 | 515 | -64% | 0 | 0 | — |
case-11 | fail→pass | 11,431 | 10,634 | -7% | 1 | 1 | 0% | 1,911 | 1,900 | -1% | 0 | 0 | — |
case-12 | pass→pass | 18,406 | 18,948 | +3% | 1 | 1 | 0% | 2,513 | 2,690 | +7% | 0 | 0 | — |
case-13 | pass→fail | 13,863 | 10,152 | -27% | 1 | 1 | 0% | 2,572 | 477 | -81% | 0 | 0 | — |
case-14 | pass→pass | 13,513 | 6,132 | -55% | 1 | 1 | 0% | 1,365 | 1,065 | -22% | 0 | 0 | — |
case-15 | pass→fail | 10,794 | 9,721 | -10% | 1 | 1 | 0% | 1,775 | 390 | -78% | 0 | 0 | — |
case-16 | pass→pass | 10,564 | 19,285 | +83% | 1 | 1 | 0% | 1,712 | 1,098 | -36% | 0 | 0 | — |
case-17 | pass→fail | 11,057 | 11,070 | +0% | 1 | 1 | 0% | 1,706 | 578 | -66% | 0 | 0 | — |
case-18 | fail→pass | 31,150 | 9,513 | -69% | 1 | 1 | 0% | 6,449 | 979 | -85% | 0 | 0 | — |
case-20 | fail→fail | 14,308 | 10,512 | -27% | 1 | 1 | 0% | 2,259 | 478 | -79% | 0 | 0 | — |
case-21 | fail→pass | 14,793 | 1,965 | -87% | 1 | 1 | 0% | 1,630 | 485 | -70% | 0 | 0 | — |
case-22 | fail→pass | 14,902 | 11,723 | -21% | 1 | 1 | 0% | 1,655 | 1,042 | -37% | 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 13 counted toward the lift figure. The other 9 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 0 percentage points is the difference between those two pass rates over the 13 comparable cases. 6 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.