| name | ideation-engine |
| description | Generate prioritized product ideas from a problem statement and score them with RICE. Use when someone says brainstorm, generate ideas, what should we build, prioritize concepts, or explore adjacent bets. Produce 10 ideas by default, support divergent, convergent, contrarian, and adjacent modes, and expand a selected idea into stories, objections, and scope.
|
| license | Apache-2.0 |
| metadata | {"author":"pm-pilot","version":"1.0.0","tags":["product-management","brainstorming","prioritization","rice","ideation"]} |
Ideation Engine
When to Use This Skill
Use this skill when the user wants solution options, prioritization support, or exploration of a product opportunity.
Read references/rice-framework.md and references/brainstorm-modes.md before generating ideas.
Default to divergent mode if the user does not specify a mode.
Instructions
- Restate the problem to anchor the session.
- Choose the brainstorming mode:
divergent, convergent, contrarian, or adjacent.
Default to divergent if the user does not specify a mode.
- Generate exactly 10 ideas unless the user requests a different count.
- For each idea, provide:
title, summary, target user, key assumption, and RICE score.
- Score ideas with explicit components:
Reach, Impact, Confidence, Effort, and computed priority.
- Do not fabricate precision.
Use relative estimates when exact numbers are unavailable.
- In
divergent mode, maximize variety but collapse near-duplicate ideas into one core concept with variants when they differ only by wording, channel, or delivery mechanism.
- Sort ideas from strongest to weakest by RICE score unless the user requests another ordering.
When ideas are tied or nearly tied, break ties in this order:
higher Confidence, lower Effort, better fit to the stated user/problem, then original order.
If the scores are still effectively tied, label them as a tie instead of inventing more precision.
- If the user selects one idea for a deep dive, expand it with:
user stories, likely objections, dependency notes, and a rough scope estimate.
- Label whether a deep-dive idea is likely
small, medium, or large.
- Keep the deep-dive behavior scoped to the selected idea only; do not replace the scored list with the expansion.
Output Format
Problem Frame
- Problem:
- Primary user:
- Mode:
- Constraints:
Idea Table
| Rank | Idea | Summary | Reach | Impact | Confidence | Effort | RICE |
|---|
If two ideas are close enough to share a rank, show the rank as a tie or shared position and explain the tie-break in the detail notes.
Idea Details
For each idea:
- Idea:
- Target user:
- Why it could work:
- Key risk:
Recommended Next Bets
- Top 3 ideas with one-line rationale each
Deep Dive
Only include this section when the user selects an idea.
- Chosen idea:
- User stories:
- Objections:
- Scope estimate:
- Suggested first release boundary:
Examples
Example Input
Brainstorm ways to reduce failed checkout recovery for subscription customers.
Example Output
Generate 10 ideas in divergent mode, score them with RICE, recommend the top three, and be ready to expand the chosen concept into stories, objections, and scope.