Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Surface Claude's assumptions about a phase approach before planning
.claude/skills/davepoon-gsd-list-phase-assumptions/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-01 | ✓→✗ | ▼ Worse | -80% | 0% |
| case-02 | ✓→✗ | ▼ Worse | -74% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -83% | 0% |
<objective> Analyze a phase and present Claude's assumptions about technical approach, implementation order, scope boundaries, risk areas, and dependencies.
Purpose: Help users see what Claude thinks BEFORE planning begins - enabling course correction early when assumptions are wrong. Output: Conversational output only (no file creation) - ends with "What do you think?" prompt </objective>
<execution_context> @${CLAUDE_PLUGIN_ROOT}/workflows/list-phase-assumptions.md </execution_context>
<context> Phase number: $ARGUMENTS (required)
Project state and roadmap are loaded in-workflow using targeted reads. </context>
<process>
</process>
<success_criteria>
</success_criteria>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-10 | pass→pass | 8,327 | 7,573 | -9% | 1 | 1 | 0% | 1,355 | 1,550 | +14% | 0 | 0 | — |
case-05 | fail→fail | 3,399 | 6,580 | +94% | 1 | 1 | 0% | 445 | 603 | +36% | 0 | 0 | — |
case-01 | pass→fail | 13,060 | 3,492 | -73% | 1 | 1 | 0% | 2,287 | 464 | -80% | 0 | 0 | — |
case-02 | pass→fail | 10,589 | 3,670 | -65% | 1 | 1 | 0% | 1,775 | 465 | -74% | 0 | 0 | — |
case-03 | fail→fail | 10,461 | 2,505 | -76% | 1 | 1 | 0% | 1,812 | 474 | -74% | 0 | 0 | — |
case-04 | pass→fail | 22,553 | 6,114 | -73% | 1 | 1 | 0% | 3,774 | 625 | -83% | 0 | 0 | — |
case-06 | pass→fail | 19,041 | 16,213 | -15% | 1 | 1 | 0% | 4,553 | 2,869 | -37% | 0 | 0 | — |
case-07 | fail→pass | 10,395 | 4,864 | -53% | 1 | 1 | 0% | 1,756 | 791 | -55% | 0 | 0 | — |
case-08 | fail→pass | 9,024 | 1,972 | -78% | 1 | 1 | 0% | 1,273 | 572 | -55% | 0 | 0 | — |
case-09 | pass→fail | 8,532 | 4,170 | -51% | 1 | 1 | 0% | 1,664 | 458 | -72% | 0 | 0 | — |
case-11 | pass→fail | 9,313 | 5,385 | -42% | 1 | 1 | 0% | 1,526 | 581 | -62% | 0 | 0 | — |
case-12 | pass→fail | 14,107 | 4,426 | -69% | 1 | 1 | 0% | 2,201 | 529 | -76% | 0 | 0 | — |
case-13 | pass→fail | 14,174 | 5,189 | -63% | 1 | 1 | 0% | 2,322 | 490 | -79% | 0 | 0 | — |
case-14 | pass→fail | 19,892 | 3,090 | -84% | 1 | 1 | 0% | 3,403 | 525 | -85% | 0 | 0 | — |
case-15 | fail→fail | 7,041 | 4,059 | -42% | 1 | 1 | 0% | 1,209 | 476 | -61% | 0 | 0 | — |
case-16 | pass→fail | 20,375 | 5,574 | -73% | 1 | 1 | 0% | 3,545 | 633 | -82% | 0 | 0 | — |
case-17 | pass→pass | 13,105 | 14,632 | +12% | 1 | 1 | 0% | 2,236 | 2,363 | +6% | 0 | 0 | — |
case-18 | pass→fail | 8,993 | 3,350 | -63% | 1 | 1 | 0% | 1,523 | 421 | -72% | 0 | 0 | — |
case-19 | fail→fail | 11,864 | 5,348 | -55% | 1 | 1 | 0% | 1,945 | 478 | -75% | 0 | 0 | — |
case-20 | pass→fail | 15,122 | 4,662 | -69% | 1 | 1 | 0% | 2,432 | 539 | -78% | 0 | 0 | — |
case-21 | pass→fail | 18,576 | 4,597 | -75% | 1 | 1 | 0% | 3,216 | 549 | -83% | 0 | 0 | — |
case-22 | pass→fail | 11,860 | 4,662 | -61% | 1 | 1 | 0% | 1,945 | 504 | -74% | 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 5 counted toward the lift figure. The other 17 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 -55 percentage points is the difference between those two pass rates over the 5 comparable cases. 15 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.