component-comparison

Prompt file imported from ktanino10/ai-hardware-engineering-team (.github/prompts/component-comparison.prompt.md). Fill in {{need}} before use. Copyright stays with the author.

Act as the Component Engineer (.github/agents/component-engineer.agent.md), following .github/skills/component-selection/SKILL.md.

Part-level need: {{need}}

Requirements context: read requirements/requirements.md for the driving requirement IDs and constraints (cost, temperature range, size, schedule).

Do:

1. Identify at least 3 candidates (fewer only with an explicit documented reason). Independent candidate research can be parallelized.
2. For each candidate, retrieve its manufacturer datasheet, register a metadata record in datasheets/ (never the actual PDF — see datasheets/README.md), and run .github/skills/datasheet-analysis/SKILL.md to extract constraints with Evidence IDs into datasheets/evidence-log.md.
3. Build the comparison table in bom/component-selection.md: electrical specs, package, price, lifecycle/EOL, availability, reference design, SDK/sample-code/docs ecosystem.
4. Recommend the candidate that maximizes project success probability (not necessarily peak spec), with explicit trade-offs and any UNKNOWNs.
5. If no datasheet can be found for a serious candidate, or this is an architecture-defining/major component decision, stop and flag it for human approval instead of proceeding.

Output: an updated bom/component-selection.md section for this need, plus a short summary of the recommendation and what (if anything) needs my approval.