Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Write a PRD for a physical hardware product — target cost (BOM and landed), industrial design constraints, regulatory certifications, reliability targets, serviceability, packaging, and forecast assumptions. Use when asked to write a hardware PRD, spec a new device, define requirements for a physical product, or kick off an NPI program. Produces a complete hardware PRD with a cost stack, cert matrix, reliability spec, and EVT/DVT/PVT milestone targets.
.claude/skills/mohitagw15856-hardware-prd/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 344% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 75% | 0% |
A software PRD can ship wrong and patch later; a hardware PRD is cut into steel and soldered onto boards. This skill writes a PRD that forces the decisions that get expensive after tooling: cost targets with a full stack (not just BOM), certifications named per market, reliability as numbers, and forecast assumptions stated so the supply chain can be sized honestly.
Ask for these if not provided; if the brief is thin, infer sensible values and label each one [assumed — confirm]:
Cost stack — always build all four layers; never let "cost target" mean BOM alone:
| Layer | Contains | |---|---| | BOM | Components, PCBA, mechanicals, packaging materials | | COGS | BOM + assembly/test labour + scrap/yield loss + factory overhead | | Landed | COGS + freight + duty/tariff (HTS code, country of origin) + insurance | | Margin check | Landed vs channel price after retailer margin (typ. 30–50% retail) |
Cert matrix — map each market to its certs. Typical starters: FCC Part 15B (unintentional) / 15C (Wi-Fi/BT radio), CE (RED, EMC, LVD), UKCA, safety UL/IEC 62368-1, RoHS/REACH/Prop 65, batteries UN 38.3 + IEC 62133. Flag any radio module choice that lacks modular certification — it adds months.
Reliability targets — state as numbers: MTBF (hours), warranty return rate ceiling (% in 12 months), lifetime (years/cycles), and the abuse cases (1 m drop, IP rating, operating temp range).
[assumed — confirm]| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 31,585 | 33,207 | +5% | 1 | 1 | 0% | 6,271 | 7,387 | +18% | 0 | 0 | — |
case-02 | fail→pass | 34,144 | 34,551 | +1% | 1 | 1 | 0% | 6,268 | 7,384 | +18% | 0 | 0 | — |
case-03 | fail→pass | 34,346 | 34,866 | +2% | 1 | 1 | 0% | 6,270 | 7,386 | +18% | 0 | 0 | — |
case-04 | fail→pass | 44,544 | 22,142 | -50% | 1 | 1 | 0% | 1,435 | 6,374 | +344% | 0 | 0 | — |
case-05 | pass→pass | 19,572 | 18,170 | -7% | 1 | 1 | 0% | 3,361 | 4,184 | +24% | 0 | 0 | — |
case-06 | pass→pass | 16,214 | 22,359 | +38% | 1 | 1 | 0% | 3,463 | 5,508 | +59% | 0 | 0 | — |
case-07 | pass→pass | 12,312 | 11,995 | -3% | 1 | 1 | 0% | 2,258 | 3,520 | +56% | 0 | 0 | — |
case-08 | pass→pass | 17,809 | 23,495 | +32% | 1 | 1 | 0% | 3,401 | 5,733 | +69% | 0 | 0 | — |
case-09 | pass→pass | 16,101 | 16,266 | +1% | 1 | 1 | 0% | 3,176 | 4,322 | +36% | 0 | 0 | — |
case-10 | pass→pass | 14,668 | 14,762 | +1% | 1 | 1 | 0% | 2,894 | 3,936 | +36% | 0 | 0 | — |
case-11 | fail→pass | 18,752 | 26,808 | +43% | 1 | 1 | 0% | 3,369 | 5,903 | +75% | 0 | 0 | — |
case-12 | pass→pass | 21,293 | 14,529 | -32% | 1 | 1 | 0% | 3,221 | 3,268 | +1% | 0 | 0 | — |
case-13 | pass→pass | 18,366 | 24,666 | +34% | 1 | 1 | 0% | 3,219 | 5,142 | +60% | 0 | 0 | — |
case-14 | pass→pass | 14,646 | 13,802 | -6% | 1 | 1 | 0% | 2,510 | 3,216 | +28% | 0 | 0 | — |
case-15 | pass→fail | 15,097 | 20,184 | +34% | 1 | 1 | 0% | 2,476 | 4,499 | +82% | 0 | 0 | — |
case-16 | pass→pass | 15,381 | 15,642 | +2% | 1 | 1 | 0% | 2,614 | 3,614 | +38% | 0 | 0 | — |
case-17 | pass→pass | 14,351 | 11,001 | -23% | 1 | 1 | 0% | 2,239 | 2,819 | +26% | 0 | 0 | — |
case-18 | fail→pass | 14,266 | 13,893 | -3% | 1 | 1 | 0% | 2,310 | 3,576 | +55% | 0 | 0 | — |
case-19 | pass→pass | 19,556 | 17,834 | -9% | 1 | 1 | 0% | 3,455 | 4,244 | +23% | 0 | 0 | — |
case-20 | pass→pass | 12,043 | 11,298 | -6% | 1 | 1 | 0% | 2,309 | 3,283 | +42% | 0 | 0 | — |
case-21 | pass→pass | 15,843 | 19,674 | +24% | 1 | 1 | 0% | 2,554 | 4,362 | +71% | 0 | 0 | — |
case-22 | fail→pass | 5,621 | 6,447 | +15% | 1 | 1 | 0% | 1,005 | 2,108 | +110% | 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 21 counted toward the lift figure. The other 1 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 21 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.