Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create a new engagement workspace. Usage: /new <platform> <program> [--type web-app|api|mobile|smart-contract]
.claude/skills/new-d32a96/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -40% | 0% |
Create a new engagement workspace: $ARGUMENTS
Parse arguments as: <platform> <program> --type <template>] Default template: web-app
Run: uv run python3 $CLAUDE_PROJECT_DIR/tools/scaffold.py $ARGUMENTS
After creation, tell the user to: cd into the workspace, start claude, and run /sync.
A new workspace is not ready until it can support evidence-grade hunting.
After scaffold, verify or create:
scope.yaml placeholder with platform, program, in-scope and out-of-scope sectionspolicy.md placeholder for required headers, rate limits, account rules, and prohibited actionsrecon/, findings/, evidence/, poc/, reports/drafts/, and scans//brain init/sync, /pipeline, /analyze, then /hunt or /autopilotWarn if the program type needs special setup: mobile test device, API tokens, sandbox tenant, OAuth app, webhook receiver, cloud account, or smart-contract fork.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,754 | 5,299 | -22% | 1 | 1 | 0% | 1,131 | 531 | -53% | 0 | 0 | — |
case-02 | fail→fail | 6,699 | 4,415 | -34% | 1 | 1 | 0% | 1,077 | 518 | -52% | 0 | 0 | — |
case-03 | fail→pass | 18,051 | 12,742 | -29% | 1 | 1 | 0% | 1,935 | 1,844 | -5% | 0 | 0 | — |
case-04 | fail→pass | 8,775 | 4,564 | -48% | 1 | 1 | 0% | 1,683 | 1,093 | -35% | 0 | 0 | — |
case-05 | fail→fail | 7,919 | 2,962 | -63% | 1 | 1 | 0% | 1,286 | 686 | -47% | 0 | 0 | — |
case-06 | fail→pass | 8,850 | 3,490 | -61% | 1 | 1 | 0% | 1,410 | 800 | -43% | 0 | 0 | — |
case-07 | fail→fail | 7,538 | 7,131 | -5% | 1 | 1 | 0% | 1,138 | 770 | -32% | 0 | 0 | — |
case-08 | fail→fail | 14,777 | 5,231 | -65% | 1 | 1 | 0% | 2,484 | 531 | -79% | 0 | 0 | — |
case-09 | fail→fail | 8,317 | 7,823 | -6% | 1 | 1 | 0% | 1,313 | 854 | -35% | 0 | 0 | — |
case-10 | fail→pass | 11,529 | 5,071 | -56% | 1 | 1 | 0% | 2,060 | 1,127 | -45% | 0 | 0 | — |
case-11 | fail→pass | 11,422 | 5,357 | -53% | 1 | 1 | 0% | 2,015 | 1,207 | -40% | 0 | 0 | — |
case-12 | fail→pass | 13,079 | 11,433 | -13% | 1 | 1 | 0% | 2,307 | 2,368 | +3% | 0 | 0 | — |
case-13 | fail→pass | 12,758 | 7,536 | -41% | 1 | 1 | 0% | 2,088 | 1,493 | -28% | 0 | 0 | — |
case-14 | fail→pass | 10,461 | 2,898 | -72% | 1 | 1 | 0% | 1,674 | 710 | -58% | 0 | 0 | — |
case-15 | fail→fail | 16,496 | 16,366 | -1% | 1 | 1 | 0% | 1,335 | 2,055 | +54% | 0 | 0 | — |
case-16 | fail→fail | 15,048 | 17,614 | +17% | 1 | 1 | 0% | 2,598 | 2,000 | -23% | 0 | 0 | — |
case-17 | fail→fail | 12,778 | 9,060 | -29% | 1 | 1 | 0% | 2,052 | 962 | -53% | 0 | 0 | — |
case-18 | fail→fail | 16,721 | 7,746 | -54% | 1 | 1 | 0% | 2,719 | 702 | -74% | 0 | 0 | — |
case-19 | fail→pass | 11,861 | 16,879 | +42% | 1 | 1 | 0% | 1,830 | 1,727 | -6% | 0 | 0 | — |
case-20 | fail→pass | 13,269 | 2,968 | -78% | 1 | 1 | 0% | 2,073 | 777 | -63% | 0 | 0 | — |
case-21 | fail→fail | 14,501 | 9,955 | -31% | 1 | 1 | 0% | 2,153 | 2,057 | -4% | 0 | 0 | — |
case-22 | fail→fail | 13,347 | 14,551 | +9% | 1 | 1 | 0% | 2,009 | 2,577 | +28% | 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 +45 percentage points is the difference between those two pass rates over the 16 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.