Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when taking over an existing codebase that has no living document. Produces a factual record of what already exists.
.claude/skills/hashgraph-online-onboard/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-18 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 1181% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -45% | 0% |
Goal: read the codebase, produce living documentation. No code changes. No plans. No gap analysis.
README.md : project overview, what it is, who it's for, how to use it.docs/tech-spec.md : main technical specification (core backbone, ≤300 lines).docs/specs/*.md : modular subsystem details split out when bulky (>15 lines). Referenced by path.Technical specification is declarations only: current ground truth facts (atemporal, no narrative, no history, no plans).
docs/tech-spec.md using the canonical format (field meanings: archive skill, "tech-spec format"):purpose / user / use-case / architecture / stack / entry /
contract / flow / invariant / constraint / conventionmilestone — onboard makes no plans.docs/specs/<topic>.md.<gate>docs/tech-spec.md must contain: stack + at least one contract + at least one convention (covering quality baseline — lint/format/typecheck tools, error-handling, security — read from code; see ../references/quality.md) before done.</gate>
docs/specs/.docs/tech-spec.md on disk → user continues with normal design → plan → tdd.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-18 | fail→pass | 21,756 | 11,255 | -48% | 1 | 1 | 0% | 2,431 | 1,541 | -37% | 0 | 0 | — |
case-01 | fail→fail | 8,754 | 13,160 | +50% | 1 | 1 | 0% | 151 | 724 | +379% | 0 | 0 | — |
case-02 | fail→fail | 8,675 | 5,414 | -38% | 1 | 1 | 0% | 240 | 848 | +253% | 0 | 0 | — |
case-03 | fail→fail | 40,112 | 11,333 | -72% | 1 | 1 | 0% | 8,240 | 862 | -90% | 0 | 0 | — |
case-04 | fail→fail | 24,458 | 10,159 | -58% | 1 | 1 | 0% | 4,949 | 905 | -82% | 0 | 0 | — |
case-05 | fail→pass | 7,832 | 15,443 | +97% | 1 | 1 | 0% | 183 | 2,345 | +1181% | 0 | 0 | — |
case-06 | fail→fail | 8,980 | 11,027 | +23% | 1 | 1 | 0% | 156 | 1,004 | +544% | 0 | 0 | — |
case-07 | fail→pass | 10,384 | 8,242 | -21% | 1 | 1 | 0% | 1,708 | 962 | -44% | 0 | 0 | — |
case-08 | fail→fail | 8,066 | 2,142 | -73% | 1 | 1 | 0% | 1,452 | 912 | -37% | 0 | 0 | — |
case-09 | fail→pass | 11,778 | 8,112 | -31% | 1 | 1 | 0% | 2,002 | 1,003 | -50% | 0 | 0 | — |
case-10 | pass→fail | 4,509 | 7,375 | +64% | 1 | 1 | 0% | 758 | 893 | +18% | 0 | 0 | — |
case-11 | fail→pass | 12,384 | 3,905 | -68% | 1 | 1 | 0% | 2,140 | 1,181 | -45% | 0 | 0 | — |
case-12 | pass→fail | 14,459 | 5,216 | -64% | 1 | 1 | 0% | 2,295 | 1,258 | -45% | 0 | 0 | — |
case-13 | fail→pass | 17,612 | 4,262 | -76% | 1 | 1 | 0% | 2,112 | 1,038 | -51% | 0 | 0 | — |
case-14 | fail→pass | 11,303 | 2,063 | -82% | 1 | 1 | 0% | 1,823 | 791 | -57% | 0 | 0 | — |
case-15 | pass→pass | 8,367 | 2,365 | -72% | 1 | 1 | 0% | 1,406 | 826 | -41% | 0 | 0 | — |
case-16 | pass→pass | 9,587 | 4,018 | -58% | 1 | 1 | 0% | 1,614 | 1,138 | -29% | 0 | 0 | — |
case-17 | fail→pass | 17,305 | 3,404 | -80% | 1 | 1 | 0% | 2,216 | 1,089 | -51% | 0 | 0 | — |
case-19 | fail→pass | 26,909 | 7,321 | -73% | 1 | 1 | 0% | 1,217 | 852 | -30% | 0 | 0 | — |
case-20 | pass→fail | 10,188 | 7,399 | -27% | 1 | 1 | 0% | 1,595 | 853 | -47% | 0 | 0 | — |
case-21 | pass→fail | 16,137 | 2,171 | -87% | 1 | 1 | 0% | 1,773 | 796 | -55% | 0 | 0 | — |
case-22 | fail→fail | 12,128 | 2,300 | -81% | 1 | 1 | 0% | 1,175 | 813 | -31% | 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 15 counted toward the lift figure. The other 7 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 +20 percentage points is the difference between those two pass rates over the 15 comparable cases. 5 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.