Ideating

Three activities with very different footprints: quick capture (zero friction), dialogue-driven ideation (expand thinking), and idea review (decide what to pursue). Match the activity to what the user brings — and keep capture frictionless above all.

Skip this skill when the user already knows what they want (understanding-the-ask) or is evaluating a specific solution (solutioning).

Activity 1 — Quick capture

The user has a thought and wants it preserved before it vanishes.

1. Accept whatever is given — a phrase, a sentence, a question.
2. Append to the project's ideas.md with the date and a one-line summary (format below).
3. Confirm saved, move on.

No forced formality: if the user says "quick idea: versioned skills", write exactly that. Don't expand, structure, or ask clarifying questions unless the user wants dialogue.

Activity 2 — Dialogue-driven ideation

Frame the exploration first — "What are we exploring? Starting fresh, or building on something?" A rough focus is enough; it doesn't need to be a problem statement.

Pick facilitation techniques to fit, and use them naturally without naming them ("What if we looked at this from a purely emotional perspective?", not "Now we're using Red Hat thinking"):

• Expansive exploration — SCAMPER prompts (substitute? combine? adapt?), analogies from other domains, random connections
• Perspective-shifting — Six Thinking Hats lenses, stakeholder viewpoints, "How might we" questions
• Assumption-surfacing — "What are we assuming must be true?", reverse an assumption and explore what becomes possible

Your highest value is introducing new dimensions, not elaborating what's already there:

• "What domain completely unrelated to this solves a similar challenge?"
• "What would someone with opposite expertise notice that we're missing?"
• "What becomes possible if we reverse [key constraint]?"

Keep divergence and convergence separate, and signal transitions explicitly: "we're generating possibilities now — no evaluation yet" → "let's shift to the critical lens". Complete divergence fully before converging (look for patterns, identify strongest directions, apply constraints, premortem).

Watch for: premature filtering ("let's capture that even if it seems wild"), anchoring to the first idea ("what would option B look like?"), surface-level responses ("and one level deeper on that?").

Capture without disrupting flow: brief in-session notes ("I'm noting: [idea]") while the user keeps talking, then offer to structure into the ideas file afterwards.

Activity 3 — Idea review

1. Read ideas.md (or wherever ideas live in the current project).
2. Summarise what exists: group by theme, note recency, highlight ideas that reference each other.
3. Facilitate selection: "Which still feels energising? Which solves the most important problem right now? Which could you prototype in a day?"
4. When the user commits to one, hand off to understanding-the-ask to clarify it into an actionable task.

Idea storage format

ideas.md in project root; chronological, append-only; human-readable and Claude-parseable. Low structure at capture, optional enrichment later.

## 2026-02-07 — Brief idea description

[Any context]

---

After an ideation session, enrich entries if the user wants: context, core concept, key insights, related ideas, optional status (Captured | Exploring | Committed | Parked).

When the user is stuck

• "I don't know what I don't know" — map stakeholders with different viewpoints, role-play each, ask what each would notice
• "I'm stuck on one approach" — "if we couldn't do that, what would we try?", reverse key constraints, analogies from distant domains
• "Everything seems possible" — must-haves vs nice-to-haves, premortem, simplest thing that could work, what could be tested in a day
• "This feels too safe" — "what would the opposite approach look like?", unlimited resource / no resource, "what assumption, if wrong, changes everything?"