Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Turn any codebase into evidence-grounded Markdown docs plus a machine-readable index.json. Every claim cites its source; never invents deployment steps.
.claude/skills/davepoon-tracedocs/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 3 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 2127% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -58% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -7% | 0% |
Turn any codebase into an evidence-grounded documentation package (overview, operation, deployment, learning, architecture, API/data, troubleshooting, maintenance) plus a machine-readable index.json for AI agents. Every operational/deployment claim cites a source file and a confidence label (Verified / Inferred / Unknown / Needs confirmation); it never invents deployment steps and records gaps instead.
Full skill, references, templates, and a validated sample output: https://github.com/wxggzz/tracedocs (MIT).
index.json)signals, tests).
confidence labels.
index.json manifest; never inventsdeployment steps.
documented).
Use tracedocs to generate evidence-grounded study docs for this repository. Write the output to study-docs/.User: "Document ./my-app with tracedocs"
The skill scans the repo and writes a study-docs/ package (00-10 manuals + index.json + _evidence/), citing each operational claim's source and labelling its confidence - and explicitly noting anything it cannot verify (for example, "no deployment configuration found in the repo").
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-08 | pass→pass | 9,166 | 5,019 | -45% | 1 | 1 | 0% | 1,526 | 1,291 | -15% | 0 | 0 | — |
case-01 | fail→fail | 28,959 | 2,833 | -90% | 1 | 1 | 0% | 5,995 | 766 | -87% | 0 | 0 | — |
case-07 | fail→fail | 9,993 | 3,415 | -66% | 1 | 1 | 0% | 1,638 | 985 | -40% | 0 | 0 | — |
case-02 | fail→pass | 4,840 | 25,071 | +418% | 1 | 1 | 0% | 243 | 5,411 | +2127% | 0 | 0 | — |
case-03 | fail→fail | 4,073 | 4,281 | +5% | 1 | 1 | 0% | 312 | 1,180 | +278% | 0 | 0 | — |
case-04 | fail→fail | 2,553 | 3,939 | +54% | 1 | 1 | 0% | 256 | 911 | +256% | 0 | 0 | — |
case-05 | pass→pass | 14,570 | 9,964 | -32% | 1 | 1 | 0% | 3,277 | 2,582 | -21% | 0 | 0 | — |
case-06 | pass→pass | 7,226 | 7,416 | +3% | 1 | 1 | 0% | 1,486 | 1,613 | +9% | 0 | 0 | — |
case-09 | fail→pass | 10,395 | 4,263 | -59% | 1 | 1 | 0% | 1,676 | 1,075 | -36% | 0 | 0 | — |
case-10 | fail→pass | 13,195 | 4,002 | -70% | 1 | 1 | 0% | 2,296 | 972 | -58% | 0 | 0 | — |
case-11 | pass→pass | 9,777 | 6,303 | -36% | 1 | 1 | 0% | 1,723 | 1,550 | -10% | 0 | 0 | — |
case-12 | pass→pass | 9,213 | 3,267 | -65% | 1 | 1 | 0% | 1,499 | 909 | -39% | 0 | 0 | — |
case-13 | fail→pass | 15,442 | 14,942 | -3% | 1 | 1 | 0% | 2,594 | 3,053 | +18% | 0 | 0 | — |
case-14 | fail→pass | 9,772 | 6,719 | -31% | 1 | 1 | 0% | 1,592 | 1,481 | -7% | 0 | 0 | — |
case-15 | fail→fail | 7,527 | 2,991 | -60% | 1 | 1 | 0% | 1,209 | 916 | -24% | 0 | 0 | — |
case-16 | fail→pass | 9,670 | 5,312 | -45% | 1 | 1 | 0% | 1,875 | 1,192 | -36% | 0 | 0 | — |
case-17 | fail→pass | 7,221 | 4,244 | -41% | 1 | 1 | 0% | 1,191 | 1,139 | -4% | 0 | 0 | — |
case-18 | pass→pass | 8,295 | 8,609 | +4% | 1 | 1 | 0% | 1,292 | 866 | -33% | 0 | 0 | — |
case-19 | pass→pass | 13,504 | 8,924 | -34% | 1 | 1 | 0% | 2,105 | 1,586 | -25% | 0 | 0 | — |
case-20 | fail→fail | 5,254 | 2,741 | -48% | 1 | 1 | 0% | 976 | 766 | -22% | 0 | 0 | — |
case-21 | pass→fail | 12,850 | 5,322 | -59% | 1 | 1 | 0% | 2,099 | 1,332 | -37% | 0 | 0 | — |
case-22 | pass→pass | 9,767 | 5,714 | -41% | 1 | 1 | 0% | 1,696 | 1,386 | -18% | 0 | 0 | — |
case-23 | pass→pass | 8,881 | 4,711 | -47% | 1 | 1 | 0% | 1,467 | 1,079 | -26% | 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. 23 cases were attempted, and 22 counted toward the lift figure. The other 1 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 +26 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is 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.