Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Deprecated — forwards to bmad-project-context. Use when the user says "document this project" or "generate project docs"
.claude/skills/bmad-code-org-bmad-document-project/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -57% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -85% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 229% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -80% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -75% | 0% |
Tell the user two things.
First: this skill is deprecated. Generating documentation volume about a codebase made agents worse, not better — agents read code more accurately than prose describing code, and the generated set was stale on arrival. bmad-project-context owns what remains useful: a small verified block in the repo's AGENTS.md carrying what the code cannot say — required policy, conventions that differ from defaults, what running the project takes that no config file states, and known pitfalls.
Second, so they are not surprised by what they get: the deeper "explain this system, its rationale and its history" material is a different altitude and is not part of that block. It is coming as its own capability. If that is what they were after, say so plainly rather than producing a thin substitute.
Then invoke bmad-project-context with setup intent, forwarding the user's original request and any paths or documents they supplied, verbatim. It takes the workflow from here.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 17,244 | 5,713 | -67% | 1 | 1 | 0% | 2,781 | 1,186 | -57% | 0 | 0 | — |
case-02 | fail→pass | 34,274 | 5,434 | -84% | 1 | 1 | 0% | 6,763 | 1,037 | -85% | 0 | 0 | — |
case-03 | fail→pass | 5,709 | 5,257 | -8% | 1 | 1 | 0% | 287 | 943 | +229% | 0 | 0 | — |
case-04 | fail→pass | 31,248 | 5,440 | -83% | 1 | 1 | 0% | 5,386 | 1,075 | -80% | 0 | 0 | — |
case-05 | fail→pass | 60,554 | 9,262 | -85% | 1 | 1 | 0% | 5,646 | 1,411 | -75% | 0 | 0 | — |
case-06 | fail→pass | 10,940 | 5,926 | -46% | 1 | 1 | 0% | 1,588 | 1,112 | -30% | 0 | 0 | — |
case-07 | fail→pass | 15,079 | 5,478 | -64% | 1 | 1 | 0% | 2,788 | 1,028 | -63% | 0 | 0 | — |
case-08 | fail→pass | 34,363 | 5,864 | -83% | 1 | 1 | 0% | 6,454 | 1,078 | -83% | 0 | 0 | — |
case-09 | fail→pass | 21,407 | 6,205 | -71% | 1 | 1 | 0% | 3,492 | 1,158 | -67% | 0 | 0 | — |
case-10 | fail→pass | 32,128 | 6,837 | -79% | 1 | 1 | 0% | 5,787 | 1,311 | -77% | 0 | 0 | — |
case-11 | fail→pass | 43,789 | 6,828 | -84% | 1 | 1 | 0% | 290 | 1,200 | +314% | 0 | 0 | — |
case-12 | fail→pass | 21,569 | 6,686 | -69% | 1 | 1 | 0% | 3,850 | 1,384 | -64% | 0 | 0 | — |
case-13 | fail→pass | 34,710 | 7,370 | -79% | 1 | 1 | 0% | 5,747 | 1,161 | -80% | 0 | 0 | — |
case-14 | fail→pass | 18,667 | 4,562 | -76% | 1 | 1 | 0% | 3,545 | 1,010 | -72% | 0 | 0 | — |
case-15 | fail→pass | 28,818 | 5,325 | -82% | 1 | 1 | 0% | 5,127 | 1,195 | -77% | 0 | 0 | — |
case-16 | fail→pass | 24,495 | 5,151 | -79% | 1 | 1 | 0% | 4,765 | 943 | -80% | 0 | 0 | — |
case-17 | fail→pass | 23,961 | 7,605 | -68% | 1 | 1 | 0% | 4,057 | 1,319 | -67% | 0 | 0 | — |
case-18 | fail→pass | 48,915 | 8,964 | -82% | 1 | 1 | 0% | 5,269 | 1,503 | -71% | 0 | 0 | — |
case-19 | fail→pass | 26,797 | 6,367 | -76% | 1 | 1 | 0% | 4,404 | 1,201 | -73% | 0 | 0 | — |
case-20 | pass→fail | 11,320 | 5,650 | -50% | 1 | 1 | 0% | 2,354 | 1,027 | -56% | 0 | 0 | — |
case-21 | pass→fail | 6,042 | 6,322 | +5% | 1 | 1 | 0% | 1,134 | 1,157 | +2% | 0 | 0 | — |
case-22 | pass→fail | 7,486 | 6,936 | -7% | 1 | 1 | 0% | 1,208 | 1,001 | -17% | 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 20 counted toward the lift figure. The other 2 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 +70 percentage points is the difference between those two pass rates over the 20 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.