Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Spec phase. Converts user requirements into a numbered Requirements + Acceptance Criteria document saved as SPEC-{timestamp}.md. Prompts /team suggestion when 3+ requirements are detected.
.claude/skills/hashgraph-online-spec/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 34% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -16% | 0% |
CRITICAL: Run HARNESS_DIR=$(epic-harness path) first. NEVER use .harness/ in the project directory.
Write a concise spec covering:
R1, R2, ...) of concrete, testable behaviorsAC1, AC2, ...) — observable outcomes that prove each requirement is metShow the spec in digestible chunks. Get explicit approval before proceeding.
Save the approved spec to $HARNESS_DIR/specs/SPEC-{timestamp}.md using this exact format:
markdown--- status: approved created: {ISO-8601 timestamp} goal_slug: {kebab-case-goal-summary} --- # SPEC-{timestamp}: {Goal} ## Goal {One sentence} ## Scope - In: {what is included} - Out: {what is explicitly excluded} ## Requirements - R1: {concrete testable behavior} - R2: {concrete testable behavior} ## Acceptance Criteria - AC1 (R1): {observable outcome proving R1} - AC2 (R2): {observable outcome proving R2} ## Technical Notes {Constraints, dependencies, edge cases}
After saving:
epic team status./team to set up a project-specific agent team before /go."/go to start building."| Excuse | Rebuttal | What to do instead | |--------|----------|-------------------| | "It's a small change, I'll just code it" | Small changes still have wrong assumptions | Write the spec — it takes 2 minutes | | "I'll refine the spec after coding" | Spec after code is documentation, not planning | Spec first, code second | | "The user didn't give me enough detail" | Then ask — don't invent requirements | Ask focused questions, max 3 at a time |
$HARNESS_DIR/specs/SPEC-{timestamp}.mdstatus: approved| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | fail→pass | 12,498 | 7,609 | -39% | 1 | 1 | 0% | 1,274 | 1,204 | -5% | 0 | 0 | — |
case-01 | fail→fail | 24,653 | 14,691 | -40% | 1 | 1 | 0% | 3,474 | 1,004 | -71% | 0 | 0 | — |
case-02 | fail→fail | 23,123 | 9,535 | -59% | 1 | 1 | 0% | 3,236 | 977 | -70% | 0 | 0 | — |
case-03 | fail→fail | 23,198 | 12,886 | -44% | 1 | 1 | 0% | 3,398 | 2,175 | -36% | 0 | 0 | — |
case-04 | fail→pass | 14,188 | 19,655 | +39% | 1 | 1 | 0% | 2,530 | 3,384 | +34% | 0 | 0 | — |
case-05 | fail→fail | 14,476 | 6,965 | -52% | 1 | 1 | 0% | 2,364 | 1,111 | -53% | 0 | 0 | — |
case-07 | fail→fail | 21,714 | 7,338 | -66% | 1 | 1 | 0% | 3,175 | 1,381 | -57% | 0 | 0 | — |
case-08 | fail→pass | 8,609 | 12,702 | +48% | 1 | 1 | 0% | 1,670 | 2,206 | +32% | 0 | 0 | — |
case-09 | fail→fail | 19,976 | 10,650 | -47% | 1 | 1 | 0% | 2,928 | 1,103 | -62% | 0 | 0 | — |
case-10 | pass→fail | 9,889 | 5,976 | -40% | 1 | 1 | 0% | 1,537 | 1,113 | -28% | 0 | 0 | — |
case-11 | fail→pass | 11,611 | 19,035 | +64% | 1 | 1 | 0% | 1,985 | 2,688 | +35% | 0 | 0 | — |
case-12 | fail→fail | 15,488 | 12,928 | -17% | 1 | 1 | 0% | 2,201 | 1,402 | -36% | 0 | 0 | — |
case-13 | fail→pass | 9,324 | 2,703 | -71% | 1 | 1 | 0% | 1,579 | 1,327 | -16% | 0 | 0 | — |
case-14 | fail→fail | 24,654 | 18,922 | -23% | 1 | 1 | 0% | 3,575 | 1,103 | -69% | 0 | 0 | — |
case-15 | fail→pass | 24,762 | 1,829 | -93% | 1 | 1 | 0% | 1,698 | 1,069 | -37% | 0 | 0 | — |
case-16 | fail→pass | 9,597 | 1,896 | -80% | 1 | 1 | 0% | 1,720 | 1,065 | -38% | 0 | 0 | — |
case-17 | pass→pass | 11,388 | 4,548 | -60% | 1 | 1 | 0% | 1,091 | 1,258 | +15% | 0 | 0 | — |
case-18 | fail→fail | 21,282 | 11,678 | -45% | 1 | 1 | 0% | 2,980 | 1,233 | -59% | 0 | 0 | — |
case-19 | pass→pass | 12,150 | 9,905 | -18% | 1 | 1 | 0% | 2,134 | 1,674 | -22% | 0 | 0 | — |
case-20 | fail→fail | 11,029 | 5,825 | -47% | 1 | 1 | 0% | 1,781 | 1,024 | -43% | 0 | 0 | — |
case-21 | fail→pass | 40,109 | 1,962 | -95% | 1 | 1 | 0% | 2,738 | 1,138 | -58% | 0 | 0 | — |
case-22 | pass→fail | 23,536 | 11,383 | -52% | 1 | 1 | 0% | 3,488 | 1,096 | -69% | 0 | 0 | — |
case-23 | pass→fail | 8,315 | 10,644 | +28% | 1 | 1 | 0% | 1,645 | 1,001 | -39% | 0 | 0 | — |
case-24 | pass→fail | 9,255 | 16,853 | +82% | 1 | 1 | 0% | 856 | 3,165 | +270% | 0 | 0 | — |
case-25 | pass→fail | 14,674 | 14,919 | +2% | 1 | 1 | 0% | 1,948 | 1,103 | -43% | 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. 25 cases were attempted, and 12 counted toward the lift figure. The other 13 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 +12 percentage points is the difference between those two pass rates over the 12 comparable cases. 7 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.