# Ideate

> Generate, challenge, and reduce a bounded set of materially different evidence-grounded possibilities before one is selected for decision or planning. Use when the user needs credible directions, improvements, or opportunities rather than refinement of one existing proposal.

- Skill: `quantipixels/ideate` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add quantipixels/ideate`
- Raw SKILL.md: https://api.skillmd.com/api/skills/quantipixels/ideate/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: quantipixels (https://skillmd.com/u/quantipixels)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/quantipixels/ideate

---


# Ideate

Expand one grounded opportunity into a bounded, mechanism-diverse candidate set; challenge every candidate; and explain only the credible survivors. Do not choose for the user, turn an idea into requirements, or skip directly to delivery.

## 1. Ground the opportunity

Pin the problem or opportunity, intended beneficiaries, current relevant reality, constraints, known non-goals, evidence boundary, time horizon, and what would make a direction useful. Ground the suggestions in current facts; use `iwadi` and `irinse` as needed.

## 2. Generate materially different mechanisms

Generate a bounded initial set before judging so early preference does not collapse the search. Cover different causal or delivery mechanisms, not cosmetic variants, renamed features, or recombinations of the same idea.

Include a conservative extension, structural alternative, or surprising direction only when current evidence makes it credible. Do not manufacture novelty to satisfy a quota.

Give every candidate a stable short ID when downstream comparison needs one and state:

- its mechanism, not merely a feature label;
- who benefits and how;
- the evidence or constraint it uses;
- main cost, dependency, operational burden, and failure mode; and
- what would have to be true for it to be worthwhile.

Stop generating when another candidate would only rename, repackage, combine, or cosmetically vary an existing mechanism. `NO_CREDIBLE_DIRECTION` is valid when the evidence supports none.

## 3. Critique every candidate

Assess each candidate against value, distinctiveness, fit, feasibility, reversibility, operational burden, trust/data boundaries, strongest credible alternative, and decisive assumptions.

Reject duplicates, weak variants, off-strategy ideas, and candidates whose expected value does not justify their cost. State one concise rejection reason for every removed candidate.

Rank survivors only when the criteria make them genuinely comparable. Otherwise group them by trade-off or decision type. Explain why each survived, what evidence it still needs, and the material decision it would create. Novelty is not a benefit by itself.

Use `ro-wo` as needed.

## 4. Return decision-ready possibilities

Return:

```text
Opportunity and evidence boundary
Bounded candidate inventory
Rejected candidates and reasons
Ranked or trade-off-grouped survivors
Decisive assumptions and evidence gaps
NO_CREDIBLE_DIRECTION when applicable
Recommended next decision/action boundary
```

Do not select, plan, persist, or visualize the survivors unless that separate result is requested or materially required downstream.