Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Implement PRDs/specs with a mandatory precheck review before coding. Use when a user asks to implement a PRD/feature spec/requirements doc or says "implement PRD/spec". Perform a preflight review, raise questions on scope/consistency/risks, then implement after confirmation.
.claude/skills/prd-implementation-precheck/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-19 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-07 | ✓→✗ | ▼ Worse | -63% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -35% | 0% |
| case-12 | ✓→✗ | ▼ Worse | 32% | 0% |
| case-13 | ✓→✗ | ▼ Worse | -9% | 0% |
Perform a short PRD precheck, surface material issues, then implement. Ask the user only when a missing choice would materially change behavior, architecture, data, or external side effects.
Verify the PRD addresses these edge cases (mark as ⚠️ if missing):
Scan the codebase to verify PRD aligns with existing patterns:
bash# Check if PRD's proposed patterns match existing code grep -r "pattern_from_prd" src/ --include="*.ts"
markdown## PRD Precheck Report ### Summary {1-2 sentence summary of what the PRD aims to achieve} ### ✅ Covered Edge Cases - {List edge cases that are well-defined in the PRD} ### ⚠️ Missing Edge Cases | Edge Case | Category | Suggested Default | Needs Confirmation | |-----------|----------|-------------------|-------------------| | Empty list display | UX | Use existing EmptyState | No | | Concurrent edit | State | Last write wins | **Yes** | ### 🔴 Blockers - {Critical issues that must be resolved before implementation} ### 🟡 Warnings - {Non-critical issues that should be addressed} ### Questions for User 1. {Specific question about missing edge case} 2. {Specific question about ambiguous requirement} --- **Proceed as-is, or update the PRD?**
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 15,659 | 16,442 | +5% | 1 | 1 | 0% | 301 | 1,289 | +328% | 0 | 0 | — |
case-02 | fail→fail | 16,263 | 15,327 | -6% | 1 | 1 | 0% | 323 | 1,148 | +255% | 0 | 0 | — |
case-03 | fail→fail | 15,209 | 14,456 | -5% | 1 | 1 | 0% | 299 | 1,191 | +298% | 0 | 0 | — |
case-04 | fail→fail | 12,688 | 16,357 | +29% | 1 | 1 | 0% | 1,251 | 1,363 | +9% | 0 | 0 | — |
case-05 | pass→pass | 14,443 | 11,486 | -20% | 1 | 1 | 0% | 1,100 | 1,973 | +79% | 0 | 0 | — |
case-06 | fail→fail | 11,847 | 16,844 | +42% | 1 | 1 | 0% | 862 | 1,237 | +44% | 0 | 0 | — |
case-07 | pass→fail | 24,220 | 15,173 | -37% | 1 | 1 | 0% | 3,089 | 1,149 | -63% | 0 | 0 | — |
case-08 | pass→fail | 20,050 | 16,461 | -18% | 1 | 1 | 0% | 1,985 | 1,294 | -35% | 0 | 0 | — |
case-09 | pass→pass | 15,076 | 19,819 | +31% | 1 | 1 | 0% | 2,114 | 2,017 | -5% | 0 | 0 | — |
case-10 | fail→fail | 20,206 | 53,919 | +167% | 1 | 1 | 0% | 2,264 | 9,310 | +311% | 0 | 0 | — |
case-11 | pass→pass | 13,904 | 9,229 | -34% | 1 | 1 | 0% | 1,318 | 1,502 | +14% | 0 | 0 | — |
case-12 | pass→fail | 8,920 | 3,575 | -60% | 1 | 1 | 0% | 1,046 | 1,376 | +32% | 0 | 0 | — |
case-13 | pass→fail | 8,611 | 11,169 | +30% | 1 | 1 | 0% | 1,291 | 1,174 | -9% | 0 | 0 | — |
case-14 | pass→fail | 7,250 | 11,762 | +62% | 1 | 1 | 0% | 1,126 | 1,228 | +9% | 0 | 0 | — |
case-15 | fail→fail | 14,448 | 10,963 | -24% | 1 | 1 | 0% | 1,863 | 1,175 | -37% | 0 | 0 | — |
case-16 | pass→pass | 17,073 | 6,100 | -64% | 1 | 1 | 0% | 1,640 | 1,975 | +20% | 0 | 0 | — |
case-17 | pass→fail | 18,389 | 10,297 | -44% | 1 | 1 | 0% | 2,049 | 1,202 | -41% | 0 | 0 | — |
case-18 | pass→pass | 9,629 | 7,373 | -23% | 1 | 1 | 0% | 1,641 | 2,166 | +32% | 0 | 0 | — |
case-19 | fail→pass | 15,697 | 9,347 | -40% | 1 | 1 | 0% | 1,895 | 1,704 | -10% | 0 | 0 | — |
case-20 | pass→fail | 29,317 | 17,498 | -40% | 1 | 1 | 0% | 5,010 | 1,321 | -74% | 0 | 0 | — |
case-21 | pass→pass | 24,057 | 29,984 | +25% | 1 | 1 | 0% | 3,399 | 5,487 | +61% | 0 | 0 | — |
case-22 | fail→fail | 7,028 | 14,988 | +113% | 1 | 1 | 0% | 945 | 1,131 | +20% | 0 | 0 | — |
case-23 | pass→fail | 39,663 | 35,228 | -11% | 1 | 1 | 0% | 5,959 | 6,449 | +8% | 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 9 counted toward the lift figure. The other 14 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 -33 percentage points is the difference between those two pass rates over the 9 comparable cases. 10 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 7/26/2026 | — |
Other measured skills in the registry, with their headline benchmark lift.