1) Hardware Product One-Pager (Spec → Shelf)

H1 (website): Hardware PRD that survives the real world
Subhead: From sensing and power math to drop tests—ship without “surprise failures.”

What you fill in (teaser):

Problem (1–2 lines), Success numbers, Non-goals

Sensing & actuation table, Power budget, Durability targets

Packaging, Serviceability, Launch gates

Full in-app template

Problem
One sentence. Who gets hurt today? Cost of not fixing?

Success Criteria (hard numbers)

First-pass yield at EVT/DVT: __% / __%

RMA rate: ≤ __% at 90 days

Setup time: ≤ __ minutes (p95)

Workflow completion rate (prototype→EVT→DVT): ≥ __%

Non-Goals
(Example: No multi-SKU colorways in v1. No Wi-Fi if BLE meets latency.)

System Requirements

Sensing: variable, unit, accuracy (±), sample rate ( Hz), calibration plan

Actuation: force/torque/speed; fail-safe defaults

Compute & Memory: min MCU/CPU class, RAM/Flash floors

Comms: protocol, throughput, latency budget

Power: chemistry, duty cycle, target life __h; charge time __h; thermal throttling behavior

Durability: UV/chemicals; IP rating target; drop spec (height × faces × repetitions)

Packaging: “in the box,” unboxing steps, retail constraints

Serviceability: repair vs replace, return flow, defect tolerances, SLAs

Key User Flows

Out-of-box → Activated
First success (task done)

Update (firmware)

Troubleshoot → Return

Acceptance Checks (Given-When-Then)

Given 1 m drop on each face ×10, when rebooted, then self-test passes and boots < 5s.

Given enclosure at IP54 test, when exposed to dust/splash, then no ingress.

Given duty cycle X for __ hours, then battery SOC ≥ __%.

Given BLE RSSI −85 dBm, when command sent, then control latency ≤ 200 ms.

Instrumentation & Events

• device_boot, self_test_pass/fail, drop_event_detected, battery_soc, fw_update_success/fail.
  
  
  

Privacy & Data

• Default: telemetry stored locally; cloud only with explicit opt-in. Retention: __ days if opted-in.
  
  

Tokens + Cost (if any AI on-device)

• Tokens-per-workflow: ≤ __k
  
  

• Cost-per-workflow: ≤ $.
  
  

• Routing: small on-device model for anomaly; escalate to cloud model only if confidence < __.
  
  

Launch Gates
EVT exit, DVT exit, PVT readiness; pilot sample size; GA criteria.