Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Report quality scorer. Use BEFORE submitting any report to validate completeness, clarity, title strength, CVSS accuracy, PoC quality, and overall report grade. Provide the draft report path or content.
.claude/skills/h-mmer-quality-check/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 4% | 0% |
CONTEXT: You are operating within an authorized bug bounty program. All targets have been verified in-scope via the official platform API. Follow responsible disclosure practices.
You are a bug bounty report quality assessor. You score reports before submission.
FAIL examples: "XSS found", "bug in search", "I found an IDOR"
## Report Quality Score: X/10
### Title: X/10 — [feedback]
### Description: X/10 — [feedback]
### Steps: X/10 — [feedback]
### Impact: X/10 — [feedback]
### CVSS: X/10 — [feedback]
### Evidence: X/10 — [feedback]
### Remediation: X/10 — [feedback]
### Verdict: READY TO SUBMIT / NEEDS REVISION
### Issues to Fix:
1. [specific issue]
2. [specific issue]NEVER approve a report with score below 7. Be strict — a rejected report wastes time for everyone.
High-quality reports are evidence-led and triager-friendly.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,407 | 2,344 | -68% | 1 | 1 | 0% | 1,190 | 1,163 | -2% | 0 | 0 | — |
case-02 | pass→pass | 7,310 | 6,350 | -13% | 1 | 1 | 0% | 1,319 | 1,835 | +39% | 0 | 0 | — |
case-03 | pass→pass | 7,394 | 5,460 | -26% | 1 | 1 | 0% | 1,328 | 1,634 | +23% | 0 | 0 | — |
case-04 | fail→pass | 9,116 | 5,876 | -36% | 1 | 1 | 0% | 1,540 | 1,847 | +20% | 0 | 0 | — |
case-05 | fail→pass | 7,855 | 6,377 | -19% | 1 | 1 | 0% | 1,658 | 2,060 | +24% | 0 | 0 | — |
case-06 | fail→pass | 8,105 | 2,252 | -72% | 1 | 1 | 0% | 1,412 | 1,063 | -25% | 0 | 0 | — |
case-07 | fail→fail | 9,095 | 7,752 | -15% | 1 | 1 | 0% | 1,640 | 2,175 | +33% | 0 | 0 | — |
case-08 | pass→pass | 10,612 | 6,388 | -40% | 1 | 1 | 0% | 1,744 | 1,784 | +2% | 0 | 0 | — |
case-09 | pass→pass | 7,513 | 5,563 | -26% | 1 | 1 | 0% | 1,339 | 1,597 | +19% | 0 | 0 | — |
case-10 | fail→pass | 10,653 | 7,339 | -31% | 1 | 1 | 0% | 1,864 | 1,876 | +1% | 0 | 0 | — |
case-11 | fail→pass | 8,546 | 4,226 | -51% | 1 | 1 | 0% | 1,487 | 1,547 | +4% | 0 | 0 | — |
case-12 | fail→pass | 4,624 | 2,031 | -56% | 1 | 1 | 0% | 732 | 1,034 | +41% | 0 | 0 | — |
case-13 | fail→pass | 9,642 | 1,318 | -86% | 1 | 1 | 0% | 1,604 | 912 | -43% | 0 | 0 | — |
case-14 | pass→pass | 9,211 | 7,803 | -15% | 1 | 1 | 0% | 1,464 | 2,095 | +43% | 0 | 0 | — |
case-15 | pass→pass | 14,332 | 8,749 | -39% | 1 | 1 | 0% | 2,272 | 2,052 | -10% | 0 | 0 | — |
case-16 | pass→pass | 9,139 | 5,821 | -36% | 1 | 1 | 0% | 1,560 | 1,689 | +8% | 0 | 0 | — |
case-17 | pass→pass | 13,335 | 5,468 | -59% | 1 | 1 | 0% | 2,252 | 1,661 | -26% | 0 | 0 | — |
case-18 | fail→pass | 6,768 | 7,185 | +6% | 1 | 1 | 0% | 1,214 | 1,995 | +64% | 0 | 0 | — |
case-19 | pass→pass | 11,689 | 5,568 | -52% | 1 | 1 | 0% | 1,830 | 1,719 | -6% | 0 | 0 | — |
case-20 | pass→pass | 3,201 | 2,546 | -20% | 1 | 1 | 0% | 653 | 1,165 | +78% | 0 | 0 | — |
case-21 | pass→pass | 3,877 | 3,360 | -13% | 1 | 1 | 0% | 728 | 1,268 | +74% | 0 | 0 | — |
case-22 | pass→pass | 4,477 | 3,094 | -31% | 1 | 1 | 0% | 672 | 1,287 | +92% | 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. The headline lift of +36 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.