Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Ideation domain specialist: Lean Canvas assembly, SPEC decomposition list extraction, and Diverge-Converge pipeline for product proposal generation. Use during /moai brain Phase 2 (Diverge), Phase 4 (Converge), and Phase 6 (Proposal).
when_to_use
Use during /moai brain ideation: Lean Canvas assembly, SPEC decomposition-list extraction, diverge/converge brainstorming, proposal generation, and structured idea exploration.
Thin orchestrator for the Diverge-Converge ideation pipeline. Delegates creative framework execution to moai-foundation-thinking and adds artifact-shaping logic specific to the brain workflow: Lean Canvas section assembly and SPEC Decomposition List extraction.
Quick Reference
Core responsibilities:
Phase 2 (Diverge): Generate 5-15 divergent concept angles for the idea
Phase 4 (Converge): Assemble Lean Canvas into ideation.md with all 9 blocks
Phase 6 (Proposal): Produce proposal.md with SPEC Decomposition Candidates section
Competitor gap angle (what existing tools fail to do)
Adjacent market angle (related problem space)
For each angle, produce a one-sentence concept label.
Cluster related angles by affinity (max 5 clusters).
Output
In-memory concept map. NOT persisted to disk at this phase — convergence in Phase 4 determines what is written.
Language Neutrality Rules
[HARD] During divergence, do NOT anchor any angle to a specific programming language or framework. Describe capabilities, not implementations:
Correct: "real-time collaborative editing engine"
Wrong: "Node.js WebSocket server with React frontend"
Phase 4: Converge — Lean Canvas Assembly
Input
Phase 2 diverged concept map
User's original idea
Optional: brand context from .moai/project/brand/brand-voice.md
Process
Invoke moai-foundation-thinking Diverge-Converge framework (Converge phase) to reduce 5-15 angles to the single most defensible product concept.
Then assemble the Lean Canvas.
Lean Canvas — 9 Blocks
Populate each block for the converged concept. Every block MUST be present, even if sparse. Empty blocks get placeholder: [TBD — to be refined with user research].
## Lean Canvas
### Problem
[Top 3 problems this product solves for the target customer]
### Customer Segments
[Specific user personas — who has the problem most acutely?]
### Unique Value Proposition
[Single, clear, compelling message — why this over alternatives]
### Solution
[Top 3 features / capabilities that address the problems]
### Channels
[How the product reaches customers: direct, marketplace, viral, partnerships]
### Revenue Streams
[How value is monetized: subscription, freemium, per-use, enterprise license, API]
### Cost Structure
[Main cost drivers: infrastructure, people, acquisition, support]
### Key Metrics
[The numbers that tell you the product is succeeding — leading and lagging indicators]
### Unfair Advantage
[What is genuinely hard for competitors to copy? Network effect, data, brand, IP, team]
Language Neutrality in Solution Block
[HARD] The Solution block describes WHAT the product does, not HOW it is built:
Correct: "High-throughput transformation engine that processes 10K events/sec"
Wrong: "Python Pandas pipeline running on Airflow DAGs"
Output
Write ideation.md to .moai/brain/IDEA-NNN/:
# Idea: {user's original idea, verbatim}*Session: {date}*## Lean Canvas
[9 blocks as specified above]
Phase 5 Append: Critical Evaluation
After Phase 5 executes (managed by moai-foundation-thinking critical-evaluation.md), append the evaluation report to the existing ideation.md:
## Evaluation Report### Strengths
[Evidence-backed strengths from critical evaluation]
### Weaknesses
[Identified gaps, assumptions, and risks]
### First Principles Validation
[First principles breakdown per moai-foundation-thinking/modules/first-principles.md]
### Verdict
[Proceed / Proceed with caveats / Revisit / Abandon — with rationale]
Phase 6: Proposal — SPEC Decomposition List
Input
ideation.md with Lean Canvas + Evaluation Report
User's confirmation to proceed
Process
Translate the converged product concept into actionable SPEC candidates. Each candidate represents a discrete, independently-implementable unit of work.
SPEC ID Naming Convention
[HARD] SPEC domain labels MUST be generic capability terms, never technology/language names:
Correct (capability-based)
Wrong (technology-based)
SPEC-AUTH-001
SPEC-OAUTH2-001
SPEC-API-001
SPEC-FASTAPI-001
SPEC-PIPELINE-001
SPEC-AIRFLOW-001
SPEC-UI-001
SPEC-REACT-001
SPEC-DB-001
SPEC-POSTGRES-001
SPEC-NOTIFY-001
SPEC-FIREBASE-001
SPEC-SEARCH-001
SPEC-ELASTICSEARCH-001
Decomposition Heuristics
Suggest 2-10 SPEC candidates. Each candidate should:
Represent a cohesive capability boundary (not too granular, not too broad)
Be independently implementable without hard dependencies on sibling SPECs (except declared dependencies)
Represent 1-3 weeks of focused work (typical SPEC scope)
Address one of the Lean Canvas Solution blocks or a key infrastructure concern
If the idea is very small (single capability), 2-3 candidates is appropriate.
If the idea is large, suggest 7-10 candidates and note that ordering matters.
Edge Case: 0 or 1 candidates
If the idea seems too atomic for SPEC decomposition:
0 candidates: Add placeholder section: ### SPEC Decomposition Candidates with note "Idea scope is atomic — consider direct /moai plan instead of /moai brain decomposition"
1 candidate: Acceptable, no special handling required
Output
Write proposal.md to .moai/brain/IDEA-NNN/:
# Proposal: {product name or concept label}*Generated: {date} | Idea: IDEA-NNN*## Product Summary
{2-3 sentence summary derived from Lean Canvas UVP + Solution blocks}
## Target User
{From Lean Canvas Customer Segments block}
## Core Problems Solved
{From Lean Canvas Problem block, formatted as numbered list}
## Proposed Solution
{From Lean Canvas Solution block — capabilities only, no tech stack}
## SPEC Decomposition Candidates
{2-10 bullets, each matching the canonical grammar below}
- SPEC-{DOMAIN}-001: {one-line scope description}
- SPEC-{DOMAIN}-002: {one-line scope description}
...
## Recommended Execution Order
{Numbered list of SPEC IDs in dependency order, with brief rationale}
## Out of Scope (v0.1)
{Explicit exclusions deferred to later SPECs or a v0.2 phase}
## Notes
{Any caveats, open questions, or assumptions from the evaluation}
Grammar Invariant (ANCHOR)
The ### SPEC Decomposition Candidates section MUST follow this exact grammar:
- SPEC-{DOMAIN}-{NUM}: {scope}
Where:
{DOMAIN} is uppercase alphanumeric (e.g., AUTH, API, UI, DB, NOTIFY)
{NUM} is zero-padded 3 digits (e.g., 001, 002, 010)
{scope} is a plain English one-line description (no backticks, no nested lists)
One bullet per line, no sub-bullets
The /moai plan --from-brain parser uses this regex: ^- SPEC-[A-Z][A-Z0-9]+-[0-9]{3}: .+$
Any bullet NOT matching this pattern is excluded from the suggestion list (surfaced as a warning, not an error).
Works Well With
moai-foundation-thinking: Diverge-Converge, Critical Evaluation, First Principles modules
moai-domain-research: Feeds research.md content into Converge phase context
moai-domain-design-handoff: Consumes proposal.md product summary for prompt.md context section
moai-workflow-brain: Orchestrates this skill across phases 2, 4, 5 (append), and 6
Common Rationalizations
Rationalization
Reality
"The user mentioned Python, so SPEC-PYTHON-001 is clearer"
Technology names in SPEC IDs create language lock-in. Use SPEC-API-001 — the technology choice happens at /moai plan time.
"The Solution block needs a tech stack to be concrete"
Solution describes WHAT the system does. HOW is deferred to architecture phase. "Processes 10K events/sec" is concrete without naming a framework.
"5 SPEC candidates is too few for a complex idea"
Start with 5-7 high-level candidates. /moai plan will decompose each one further if needed.
"I should skip the Lean Canvas for a simple idea"
Every brain invocation produces a Lean Canvas. The Customer Segments block alone is worth the exercise — it forces explicit user definition.
Verification
Phase 2 diverge produced 5-15 distinct angles (not minor variations of the same angle)
Phase 4 Lean Canvas has all 9 blocks present (none omitted or collapsed)
Solution block contains no technology/framework names (capability-only language)
