Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Record a platform response and update learning. Usage: /learn <report_id> <status> [--bounty 500] [--vuln-type XSS]
.claude/skills/h-mmer-learn/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -72% | 0% |
Record platform response: $ARGUMENTS
uv run python3 ../../tools/response_tracker.py log $ARGUMENTSuv run python3 ../../tools/brain.py log "Report response: $ARGUMENTS"uv run python3 ../../tools/global_brain.py sync-from-localuv run python3 ../../tools/response_tracker.py insightsConvert every platform response into a future hunting rule.
End with one concrete update: a brain pattern, a never-submit rule, a report wording change, or a target ranking adjustment.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 12,834 | 19,884 | +55% | 1 | 1 | 0% | 1,620 | 2,336 | +44% | 0 | 0 | — |
case-22 | pass→pass | 10,601 | 15,370 | +45% | 1 | 1 | 0% | 1,654 | 1,899 | +15% | 0 | 0 | — |
case-02 | fail→fail | 8,886 | 17,074 | +92% | 1 | 1 | 0% | 1,349 | 567 | -58% | 0 | 0 | — |
case-03 | fail→fail | 16,849 | 4,631 | -73% | 1 | 1 | 0% | 3,088 | 509 | -84% | 0 | 0 | — |
case-04 | fail→fail | 12,081 | 6,533 | -46% | 1 | 1 | 0% | 1,873 | 515 | -73% | 0 | 0 | — |
case-05 | fail→pass | 16,925 | 19,162 | +13% | 1 | 1 | 0% | 1,580 | 2,108 | +33% | 0 | 0 | — |
case-21 | pass→pass | 12,878 | 11,016 | -14% | 1 | 1 | 0% | 2,673 | 2,462 | -8% | 0 | 0 | — |
case-06 | fail→fail | 15,968 | 5,424 | -66% | 1 | 1 | 0% | 2,619 | 533 | -80% | 0 | 0 | — |
case-07 | fail→fail | 7,160 | 4,849 | -32% | 1 | 1 | 0% | 1,223 | 549 | -55% | 0 | 0 | — |
case-08 | pass→fail | 10,420 | 4,431 | -57% | 1 | 1 | 0% | 1,678 | 470 | -72% | 0 | 0 | — |
case-09 | pass→fail | 10,926 | 5,284 | -52% | 1 | 1 | 0% | 1,795 | 477 | -73% | 0 | 0 | — |
case-10 | fail→fail | 10,555 | 4,431 | -58% | 1 | 1 | 0% | 1,721 | 520 | -70% | 0 | 0 | — |
case-11 | fail→fail | 15,000 | 4,584 | -69% | 1 | 1 | 0% | 2,447 | 457 | -81% | 0 | 0 | — |
case-12 | pass→fail | 10,852 | 4,590 | -58% | 1 | 1 | 0% | 1,799 | 457 | -75% | 0 | 0 | — |
case-13 | fail→fail | 11,067 | 5,316 | -52% | 1 | 1 | 0% | 1,792 | 561 | -69% | 0 | 0 | — |
case-14 | pass→fail | 10,440 | 5,818 | -44% | 1 | 1 | 0% | 1,578 | 667 | -58% | 0 | 0 | — |
case-15 | fail→pass | 7,777 | 4,846 | -38% | 1 | 1 | 0% | 1,439 | 1,175 | -18% | 0 | 0 | — |
case-16 | fail→pass | 9,550 | 4,868 | -49% | 1 | 1 | 0% | 1,682 | 1,160 | -31% | 0 | 0 | — |
case-17 | fail→fail | 14,290 | 5,517 | -61% | 1 | 1 | 0% | 2,673 | 619 | -77% | 0 | 0 | — |
case-18 | fail→pass | 8,528 | 6,893 | -19% | 1 | 1 | 0% | 1,498 | 1,572 | +5% | 0 | 0 | — |
case-19 | pass→fail | 6,916 | 4,963 | -28% | 1 | 1 | 0% | 1,092 | 575 | -47% | 0 | 0 | — |
case-20 | pass→pass | 10,387 | 10,306 | -1% | 1 | 1 | 0% | 1,844 | 2,051 | +11% | 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 8 counted toward the lift figure. The other 14 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 -5 percentage points is the difference between those two pass rates over the 8 comparable cases. 8 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.