Ideate

Ideate

Turns a problem statement into a small set of validated, evolved ideas by simulating a nominal group of independent generators, scoring with independent judges, and breeding the winners — instead of asking one context for a list and polishing it.

This SKILL.md is a thin index. Phase procedures live in rules/*.md and load on demand. The evidence behind every numerical default lives in references/ideation-research.md — section references below (§) point there, and its defaults tables flag which numbers are study values vs the skill's own operationalizations.

Contents

• Mode Detection
• Workflow
• Quick vs deep
• Composition
• Core Principles
• Self-Improvement
• Anti-patterns
• Definition of Done

---

Mode Detection

Parse $ARGUMENTS:

Mode	Default	Trigger
quick		quick token, or auto-triage says small.
deep		deep token, or auto-triage says open-ended.
auto	yes	No mode token — run the depth triage table below.

Flag	Default	Meaning
--n <count>	unset	Optional cap on the Lead-finalists list. Unset by default — the finalist count is emergent from the quality bar (see idea-scoring.md § Selection rules). --n only narrows; it never manufactures finalists the bar did not admit.
--no-framing	off	Skip Phase 1 (problem is already well-framed).

Everything else in $ARGUMENTS is the problem statement. If it is missing, ask for it — never ideate on a guessed problem.

Depth triage (auto mode)

First match wins:

#	Signal	Depth
1	User asks for thorough / extensive / "really new" ideas, or the problem combines ≥ 2 domains ("X meets Y").	deep
2	Outcome is a product, business, strategy, or roadmap decision.	deep
3	Problem is a local decision: naming, small feature shape, workaround, single component.	quick
4	User is mid-conversation and wants options fast ("any ideas?", "what do you think?").	quick
5	Unsure.	quick, and name the escalation path: "run /ideate deep <problem> for the full pipeline".

Cost expectation: a deep run with a typical 2–4-idea finalist band dispatches ~17 subagents (2 bursts × 5 generators, 1 pool judge, 1 breeder, 1 variant judge, 3 panel judges, 1 pre-mortem) — prefer quick for casual or budget-sensitive asks. A richer pool that admits more finalists adds only panel-judge cost, not generation cost.

---

Workflow

Phase	Name	Rule file	Gate
0	Intake & triage	(inline below)	Problem restated in one sentence; success criterion named; depth chosen; lessons read.
1	Frame	(inline below)	3–5 "How Might We" framings at different widths; one selected and logged.
2	Diverge	rules/divergence.md	Pool ≥ 12 (quick) / ≥ 25 (deep) unique ideas; zero evaluation happened during generation.
3	Score	rules/idea-scoring.md	Every pooled idea scored on 4 independent axes by a non-generator judge.
4	Evolve	rules/evolution-loop.md	≤ 3 rounds; stopped on flat external scores, never on self-assessed improvement.
5	Validate	rules/idea-scoring.md § Finalist validation	Every finalist has an executability probe; confidence(analysis) ≥ 70 on the recommendation.
6	Report	templates/ideation-report.md	Report emitted; finalist count reflects the quality bar; high-novelty wildcard included; "Ideas worth revisiting" filled; verdict question asked.
7	Learn	rules/self-improvement-loop.md	Lessons written — mechanics only, never idea content.

Phase 0 — Intake & triage

1. Read lessons (advisory input for mechanics only — see the hard invariant under Self-Improvement). Narrow-to-broad LoreKit fan-out; skips silently if memory.* is not connected:

   memory.list { scope: "repo::{owner}/{repo}", tags: ["loop::ideate-lessons"], limit: 50 }
   memory.list { scope: "global",               tags: ["loop::ideate-lessons"], limit: 50 }
   

2. Restate the problem in one sentence and name the success criterion — what makes an idea "good" here (cheapest, most novel, shippable this week, …).

3. Run the depth triage table unless a mode token was given.

4. If the user supplied seed ideas, add them to the pool unlabeled — judges must not know which ideas are the user's.

Phase 1 — Frame

Problem framing measurably shapes ideation breadth and direction (§2.5) — skip only with --no-framing.

1. Generate 3–5 "How Might We" framings at deliberately different widths, from the literal ask to the underlying job-to-be-done — the width variation is the evidenced part (§2.5); the HMW phrasing is convention.
2. Interactive session → ask the user to pick one (a single batched question). Autonomous or backgrounded → pick the framing one step wider than the literal ask, and log the choice in the report.

---

Quick vs deep

Aspect	quick	deep
Generation	In-context, 2 bursts × 8 ideas, operators rotated per burst.	5 parallel persona subagents × 6 ideas per burst, ≥ 2 bursts (§1.4, §4.4).
Fixation guard	Operator switch + "what has NOT been said yet" reseed (§1.5).	Independent contexts — the structural fix (§4.3).
Judging	Separate in-context pass with rubric, order-swapped pairwise.	Fresh judge subagent; panel of 3 judges for finalists (§4.7).
Evolution	≤ 1 round.	≤ 3 rounds (§5.1).
Report	Inline.	Inline + written to .agent/ideate/<yyyy-mm-dd>-<slug>.md.

---

Composition

Skill	When	Call
confidence	Phase 5, on the finalist recommendation.	Skill("confidence", "analysis")
critical	Deep mode, top finalist (automatic); any finalist on request.	Skill("critical", "analysis") — run inside a fresh subagent so the pre-mortem stays adversarial and the main context stays lean.
ux	Finalists that are UI/UX or product-surface concepts.	Skill("ux") as a lens on the finalist.
lorekit-memory (LoreKit memory.* tools)	Phases 0 and 7.	See rules/self-improvement-loop.md.

confidence is required; the others are optional — skip silently if not installed / not connected.

---

Core Principles

1. Simulate a nominal group, not a meeting. Independent generation contexts, pooled afterwards — never one long list from one context (§1.2, §4.3).
2. Strictly separate generation from evaluation. The one Osborn rule that survives scrutiny; no feasibility talk inside a generation pass (§3.1).
3. Diversity comes from personas and prompts, not temperature. Temperature is a weak novelty lever with a coherence cost (§4.6).
4. The judge is never the generator. Self-scoring amplifies self-bias per iteration (§4.7, §5.1).
5. Protect novelty at selection. Default selection sacrifices originality for feasibility and performs near-randomly; always carry one high-novelty pick (§3.2).
6. Gate finalists on quality, not quantity. The finalist count is emergent: report every idea that clears the admission bar, not a fixed top-3. A fixed count drops deserving ideas from a rich pool and pads a thin one (§3.2).
7. Iterate by recombination across lineages, not polishing. Cap at 3 rounds; stop on flat external scores (§5.1–§5.3).
8. The obvious ideas come first. Always run a second burst seeded with "what has NOT been said yet" (§1.5).
9. Pre-execution novelty is inflated. Every finalist needs a concrete first-step probe before it is recommended (§4.2).

---

Self-Improvement

Two-tier loop, scope ideate-lessons — full contract in rules/self-improvement-loop.md, diagnostic surface in rules/diagnostic-surface.md.

Hard invariant: divergence runs lessons-blind. Lessons may inform mechanics — depth triage, operator effectiveness, judge calibration, stopping behavior — and must never seed, filter, or steer idea content. "What kinds of ideas the user tends to pick" is on the never-store list: it would entrench homogenization, the exact failure this skill exists to avoid.

---

Anti-patterns

• One 30-idea list from a single context — within-context fixation makes the last 20 near-duplicates.
• Scoring or feasibility talk during a generation pass.
• Letting the generation context judge its own output.
• Averaging the four axes into one number before selection, then picking the top-n — this silently discards every high-novelty idea.
• Capping the report at a fixed 3–4 finalists regardless of how many ideas cleared the quality bar — the count is emergent, not a target.
• Treating the "Ideas worth revisiting" section as a throwaway list — it must name why each near-miss missed and what would flip it; that is the inspiring part.
• A 4th evolution round because it "still feels like it's improving" — self-assessed improvement is the signal that lies.
• Storing user idea-taste as a lesson.

---

Definition of Done

• Problem restated and success criterion named before any generation.
• Generation and evaluation never co-occurred in one pass.
• Pool met the unique-idea gate for the chosen depth.
• Every finalist has all four axis scores, an executability probe, and the confidence gate result.
• The finalist count reflects the quality bar, not a fixed target — every idea that cleared the admission bar is reported (or capped only by an explicit --n).
• Report includes the high-novelty wildcard, the run stats (bursts, non-duplicate yield, evolution rounds, score trajectory), and an "Ideas worth revisiting" section that names, per near-miss, why it missed and what would flip it.
• User verdict requested; lessons written per the loop contract.