project-brainstorm
Input: $ARGUMENTS (raw ideas, problems, inspirations — or empty for interactive).
Result: project concept(s) ready for /project-specify; optionally save specs/projects/brainstorm-[timestamp].md.
Position: BEFORE /project-specify — optional ideation.
1. Gather ideas
With args: parse distinct ideas; ask "Anything else before we explore?"
Without args: prompt for problems, frustrations, wished-for products, domains, technologies. Continue until enough material or user done.
2. Cluster themes
Identify: problem themes, solution themes, user themes, domain/regulatory themes. Present clusters; confirm or correct.
3. Generate concepts
Per viable theme (1–3 concepts):
Name · One-liner
Problem: [who + pain + impact]
Solution: [product type + core capability + benefit]
Users: primary + secondary
Key features: 3–5 bullets
Differentiation vs existing solutions
Complexity: Level 0–4 · timeframe · technical challenges
Confidence: High/Medium/Low + reason
Optional web search: competitors, market gap, enabling tech → append "Quick research" bullets.
4. Explore options
Offer: deep dive · combine concepts · more variations · proceed to /project-specify · continue brainstorming.
Deep dive: competitors, feasibility, risks, market (if applicable).
5. Handoff
When user selects concept, emit one-paragraph description for /project-specify:
/project-specify [description]
Ask: run now or save session?
Optional save → specs/projects/brainstorm-[timestamp].md:
- Raw ideas, themes, all concepts, selected, rejected + reasons
6. Router signals
Route here (not /project-specify) when: vague idea, "help me figure out what to build", brainstorm intent, insufficient specificity.
NEVER / ALWAYS
- NEVER jump to
/project-specifybefore user picks a concept - NEVER dismiss ideas without exploration
- NEVER omit rejected concepts from saved session (if saving)
- ALWAYS produce at least one concrete concept
- ALWAYS make handoff description specific enough to specify