Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Capture a forward-looking idea with trigger conditions — surfaces automatically at the right milestone
.claude/skills/coco-research-gsd-plant-seed/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | 45% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-06 | ✓→✗ | ▼ Worse | -60% | 0% |
| case-07 | ✓→✗ | ▼ Worse | -57% | 0% |
| case-21 | ✓→✗ | ▼ Worse | -84% | 0% |
<objective> Capture an idea that's too big for now but should surface automatically when the right milestone arrives. Seeds solve context rot: instead of a one-liner in Deferred that nobody reads, a seed preserves the full WHY, WHEN to surface, and breadcrumbs to details.
Creates: .planning/seeds/SEED-NNN-slug.md Consumed by: /gsd-new-milestone (scans seeds and presents matches) </objective>
<execution_context> @$HOME/.claude/get-shit-done/workflows/plant-seed.md </execution_context>
<process> Execute the plant-seed workflow from @$HOME/.claude/get-shit-done/workflows/plant-seed.md end-to-end. </process>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 14,538 | 14,571 | +0% | 1 | 1 | 0% | 1,380 | 438 | -68% | 0 | 0 | — |
case-02 | fail→fail | 12,339 | 11,652 | -6% | 1 | 1 | 0% | 1,888 | 492 | -74% | 0 | 0 | — |
case-03 | fail→fail | 11,924 | 15,456 | +30% | 1 | 1 | 0% | 1,850 | 870 | -53% | 0 | 0 | — |
case-04 | fail→fail | 13,325 | 13,693 | +3% | 1 | 1 | 0% | 1,347 | 577 | -57% | 0 | 0 | — |
case-05 | fail→fail | 11,271 | 12,169 | +8% | 1 | 1 | 0% | 898 | 2,274 | +153% | 0 | 0 | — |
case-18 | fail→fail | 10,400 | 15,295 | +47% | 1 | 1 | 0% | 747 | 391 | -48% | 0 | 0 | — |
case-06 | pass→fail | 9,299 | 5,918 | -36% | 1 | 1 | 0% | 1,310 | 528 | -60% | 0 | 0 | — |
case-07 | pass→fail | 13,281 | 17,909 | +35% | 1 | 1 | 0% | 1,336 | 579 | -57% | 0 | 0 | — |
case-08 | fail→fail | 20,477 | 12,924 | -37% | 1 | 1 | 0% | 2,252 | 679 | -70% | 0 | 0 | — |
case-09 | fail→fail | 19,194 | 4,655 | -76% | 1 | 1 | 0% | 2,259 | 466 | -79% | 0 | 0 | — |
case-10 | fail→pass | 24,103 | 12,362 | -49% | 1 | 1 | 0% | 1,110 | 1,604 | +45% | 0 | 0 | — |
case-11 | fail→fail | 16,932 | 16,156 | -5% | 1 | 1 | 0% | 1,669 | 426 | -74% | 0 | 0 | — |
case-12 | fail→fail | 20,191 | 11,439 | -43% | 1 | 1 | 0% | 2,685 | 533 | -80% | 0 | 0 | — |
case-13 | fail→fail | 27,478 | 17,969 | -35% | 1 | 1 | 0% | 2,945 | 358 | -88% | 0 | 0 | — |
case-14 | fail→fail | 7,165 | 15,156 | +112% | 1 | 1 | 0% | 1,239 | 478 | -61% | 0 | 0 | — |
case-15 | fail→fail | 15,201 | 24,466 | +61% | 1 | 1 | 0% | 2,297 | 3,194 | +39% | 0 | 0 | — |
case-16 | fail→fail | 10,320 | 6,336 | -39% | 1 | 1 | 0% | 885 | 388 | -56% | 0 | 0 | — |
case-17 | fail→fail | 11,808 | 17,508 | +48% | 1 | 1 | 0% | 1,737 | 447 | -74% | 0 | 0 | — |
case-19 | fail→fail | 18,223 | 9,783 | -46% | 1 | 1 | 0% | 2,114 | 337 | -84% | 0 | 0 | — |
case-20 | fail→fail | 12,039 | 16,702 | +39% | 1 | 1 | 0% | 1,947 | 414 | -79% | 0 | 0 | — |
case-21 | pass→fail | 20,017 | 17,073 | -15% | 1 | 1 | 0% | 2,751 | 439 | -84% | 0 | 0 | — |
case-22 | fail→pass | 17,848 | 24,859 | +39% | 1 | 1 | 0% | 2,167 | 2,609 | +20% | 0 | 0 | — |
case-23 | fail→fail | 14,132 | 9,850 | -30% | 1 | 1 | 0% | 2,002 | 424 | -79% | 0 | 0 | — |
case-24 | fail→fail | 16,741 | 14,783 | -12% | 1 | 1 | 0% | 2,879 | 614 | -79% | 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. 24 cases were attempted, and 3 counted toward the lift figure. The other 21 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 -4 percentage points is the difference between those two pass rates over the 3 comparable cases. 14 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.