Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Capture raw thoughts with automatic domain classification and vault routing
.claude/skills/a5c-ai-cog-braindump-capture/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -59% | 0% |
file-read to load user profile from 00-inbox for classification contextfile-write to create classified entries in appropriate vault directoriesfile-search to find related existing entries for cross-referencingfile-write to add cross-references to new and existing entriesgit-commit to commit captured contentjson{ "vaultPath": "./cog-vault", "captureType": "braindump", "content": "Had a great idea about the auth system redesign. Also need to book vacation for July. The React 19 features look promising for our dashboard project.", "targetQuality": 80 }
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 4,037 | 4,832 | +20% | 1 | 1 | 0% | 656 | 545 | -17% | 0 | 0 | — |
case-02 | fail→fail | 10,680 | 6,407 | -40% | 1 | 1 | 0% | 2,295 | 724 | -68% | 0 | 0 | — |
case-03 | fail→fail | 3,434 | 5,154 | +50% | 1 | 1 | 0% | 292 | 516 | +77% | 0 | 0 | — |
case-04 | fail→pass | 6,812 | 5,632 | -17% | 1 | 1 | 0% | 1,236 | 757 | -39% | 0 | 0 | — |
case-05 | fail→pass | 6,694 | 3,241 | -52% | 1 | 1 | 0% | 1,063 | 794 | -25% | 0 | 0 | — |
case-06 | fail→pass | 5,810 | 2,084 | -64% | 1 | 1 | 0% | 975 | 618 | -37% | 0 | 0 | — |
case-07 | fail→fail | 10,086 | 1,911 | -81% | 1 | 1 | 0% | 1,751 | 540 | -69% | 0 | 0 | — |
case-08 | pass→pass | 9,230 | 3,318 | -64% | 1 | 1 | 0% | 1,517 | 722 | -52% | 0 | 0 | — |
case-09 | fail→fail | 5,400 | 1,659 | -69% | 1 | 1 | 0% | 906 | 545 | -40% | 0 | 0 | — |
case-10 | fail→fail | 5,613 | 1,698 | -70% | 1 | 1 | 0% | 828 | 498 | -40% | 0 | 0 | — |
case-11 | fail→fail | 8,464 | 1,614 | -81% | 1 | 1 | 0% | 1,470 | 503 | -66% | 0 | 0 | — |
case-12 | pass→pass | 7,778 | 1,235 | -84% | 1 | 1 | 0% | 1,167 | 475 | -59% | 0 | 0 | — |
case-13 | fail→fail | 4,631 | 1,875 | -60% | 1 | 1 | 0% | 716 | 508 | -29% | 0 | 0 | — |
case-14 | fail→pass | 10,190 | 4,309 | -58% | 1 | 1 | 0% | 1,668 | 942 | -44% | 0 | 0 | — |
case-15 | fail→pass | 7,715 | 1,684 | -78% | 1 | 1 | 0% | 1,107 | 458 | -59% | 0 | 0 | — |
case-16 | fail→fail | 7,983 | 1,905 | -76% | 1 | 1 | 0% | 1,433 | 538 | -62% | 0 | 0 | — |
case-17 | fail→fail | 14,232 | 10,565 | -26% | 1 | 1 | 0% | 2,283 | 1,774 | -22% | 0 | 0 | — |
case-18 | fail→fail | 7,546 | 5,453 | -28% | 1 | 1 | 0% | 1,456 | 668 | -54% | 0 | 0 | — |
case-19 | fail→pass | 8,037 | 1,966 | -76% | 1 | 1 | 0% | 1,102 | 634 | -42% | 0 | 0 | — |
case-20 | fail→fail | 12,576 | 7,559 | -40% | 1 | 1 | 0% | 2,735 | 1,886 | -31% | 0 | 0 | — |
case-21 | fail→fail | 14,390 | 15,659 | +9% | 1 | 1 | 0% | 3,209 | 3,174 | -1% | 0 | 0 | — |
case-22 | fail→fail | 10,957 | 8,170 | -25% | 1 | 1 | 0% | 2,168 | 1,817 | -16% | 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 18 counted toward the lift figure. The other 4 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 +27 percentage points is the difference between those two pass rates over the 18 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.