| name | idea-explore |
| description | Deep multi-framework analysis of a problem or product concept, broader than the proposal pipeline. |
| when_to_use | Use for deep multi-lens analysis of a problem or product idea — broader than `/product-explore`'s proposal pipeline. Also when user says "explore this idea", "multiple frameworks", "different lenses", "is this idea good", "stress-test this concept", "should I pursue this", or "deep idea analysis".
|
| user-invocable | true |
| allowed-tools | Agent AskUserQuestion Read Write Edit Glob Grep Bash |
Idea Explore Pipeline
Spawns 8 expert framework analyst agents in parallel, then a team-lead synthesizer that produces the final markdown report.
Frameworks (8 agents across 4 categories)
| Category | Framework | Agent |
|---|
| 1. Reframing / problem definition | Jobs-to-be-Done | ie-jtbd |
| 1. Reframing / problem definition | Opportunity Solution Tree | ie-ost |
| 2. Idea expansion / divergent thinking | Morphological Analysis | ie-morphological |
| 2. Idea expansion / divergent thinking | Assumption Reversal | ie-assumption-reversal |
| 3. Strategic positioning | Blue Ocean (Strategy Canvas + ERRC) | ie-blue-ocean |
| 3. Strategic positioning | Wardley Mapping | ie-wardley |
| 4. Defensibility & Validation | 7 Powers (Helmer) | ie-seven-powers |
| 4. Defensibility & Validation | Mom Test / Customer Discovery | ie-mom-test |
Synthesizer: ie-team-lead.
Inputs
The skill takes free-form input and extracts:
problem (required) — description of the problem / opportunity space
product (optional) — proposed solution, product idea, or hypothesis
If the user invoked /idea-explore with no arguments, ask once: "Describe the problem you want analyzed. Optionally, also describe the product idea you have in mind." Then proceed — do not ask further clarifying questions about scope or framework methodology.
Pipeline
Step 1 — Parse input
From the user message, extract problem and product. product may be empty (analysis still works on problem-space alone).
Step 2 — Generate output slug
Derive a kebab-case slug: 3–5 meaningful words from the problem, prefixed with the date.
Format: YYYY-MM-DD-<3-5-word-slug>
Example: 2026-05-06-async-mahjong-onboarding
Step 3 — Spawn 8 framework agents IN PARALLEL
This is critical: send a single message with 8 Agent tool calls. Sequential invocation is wrong, defeats the parallel design, and triples wall-clock time.
Each agent receives the same input contract:
PROBLEM: <problem>
PRODUCT: <product or "(not specified)">
Apply your framework lens. Follow your output spec exactly.
Agents to spawn (all eight, in one message):
ie-jtbd
ie-ost
ie-morphological
ie-assumption-reversal
ie-blue-ocean
ie-wardley
ie-seven-powers
ie-mom-test
Use subagent_type matching each agent's name. Models are pinned in each agent's frontmatter — do not override.
Step 4 — Spawn team-lead synthesizer
After all 8 framework agents return, spawn ie-team-lead with:
PROBLEM: <problem>
PRODUCT: <product or "(not specified)">
SLUG: <slug from Step 2>
OUTPUT_PATH: docs/idea-explorations/<slug>.md
=== FRAMEWORK OUTPUTS ===
[1/8] JTBD:
<full output from ie-jtbd>
[2/8] Opportunity Solution Tree:
<full output from ie-ost>
[3/8] Morphological Analysis:
<full output from ie-morphological>
[4/8] Assumption Reversal:
<full output from ie-assumption-reversal>
[5/8] Blue Ocean:
<full output from ie-blue-ocean>
[6/8] Wardley Mapping:
<full output from ie-wardley>
[7/8] 7 Powers:
<full output from ie-seven-powers>
[8/8] Mom Test:
<full output from ie-mom-test>
Team lead writes the synthesized markdown to OUTPUT_PATH. Create the parent directory if missing.
Step 5 — Report to user
After team lead completes, give a short summary:
- Path to the report
- 2–3 sentence headline finding (convergence point)
- One pointed question or recommended next step
Rules
- Always parallel. The 8 framework agents MUST be invoked in a single message with 8 Agent tool calls. Never sequential.
- No clarifying questions about framework methodology — each agent owns its own approach.
- Output location:
docs/idea-explorations/<slug>.md. Create the directory if it doesn't exist.
- Don't run framework agents inside the team lead. Team lead is a synthesizer, not an orchestrator.
- Models are agent-pinned. Framework agents = sonnet/high. Team lead = opus/xhigh.
- If the user gave only a problem (no product), all 8 agents still apply — they shift to opportunity-space analysis instead of product-space analysis.
Failure modes to avoid
- Asking a stack of clarifying questions before spawning agents — the agents themselves will name the questions worth asking
- Summarizing framework outputs yourself before passing them to team lead — pass the raw outputs verbatim
- Skipping a framework because "it doesn't apply" — let the agent itself say so in its output
- Writing the markdown doc yourself — that's the team lead's job