bmad-idea（疯狂的想法）

BMAD Idea Router

Use this skill as the repository's pre-planning creative intake and concept-routing layer.

The job is not to dump dozens of ideation techniques or persona gimmicks. The job is to:

1. normalize a messy early-stage idea packet,
2. choose the single framing mode that reduces ambiguity fastest,
3. produce one concept artifact that downstream work can actually use,
4. route the next step to the right neighboring skill.

Read references/operating-modes.md for mode-selection heuristics, references/handoff-boundaries.md for adjacent-skill routing, and references/concept-packet-template.md for reusable output scaffolding.

When to use this skill

• The user has a rough idea and needs a clearer problem, audience, concept, or story before planning starts
• The request is still ambiguous enough that jumping straight to a PRD, sprint plan, or launch plan would be premature
• A product, service, content, or GTM idea needs one concept brief, positioning brief, or narrative packet first
• A game idea needs pillars, player fantasy, core loop, and first concept framing before bmad-gds or task decomposition
• The user explicitly mentions brainstorming, concept shaping, design thinking, positioning, JTBD-style framing, story packaging, or pre-planning clarification

When not to use this skill

• The main need is phase-based delivery routing or deciding the next BMAD artifact after project framing already exists → use bmad
• The main need is backlog slicing, story decomposition, estimation, or sprint preparation → use task-planning
• The main need is broad marketing execution, launch planning, funnel work, or KPI-aware campaign routing → use marketing-automation
• The main need is game-production coordination for an existing GDD, playtest packet, build issue, or milestone → use bmad-gds
• The main need is only rewriting copy or polishing prose with no framing decision to make

Operating modes

Choose exactly one primary mode per run.

1. Problem framing

- Use when the idea is fuzzy and the main gap is *what problem actually matters* - Best for founders, PMs, designers, and consultants starting from raw opportunity language

1. Audience and value framing

- Use when the offer exists but the target user, JTBD, pains, gains, or differentiation are weak - Best for product marketing, GTM, and concept-validation work

1. Concept shaping

- Use when the direction is promising but still needs a tighter concept, constraints, options, and core promise - Best for product, feature, service, and offer-level idea shaping

1. Game concept framing

- Use when the packet is about game fantasy, pillars, audience, core loop, progression, or tone before production planning - Best for solo developers and small game teams before bmad-gds

1. Story packaging

- Use when the main need is a narrative, pitch, concept story, or internal alignment packet - Best when the problem is understanding and buy-in rather than raw ideation volume

Instructions

Step 1: Normalize the idea intake

Capture the packet in this format before choosing a mode:

idea_packet:
  domain: product | gtm | content | consulting | game | mixed | unknown
  stage: raw-idea | rough-concept | candidate-offer | concept-with-notes | narrative-draft | unknown
  audience:
    primary_user: "who this is for"
    pains_or_needs:
      - pain 1
      - pain 2
  problem_or_goal: "what change or outcome is this idea trying to create?"
  current_material:
    - notes
    - brainstorm bullets
    - positioning draft
    - customer research
    - game concept notes
    - moodboard / references
    - pitch / story draft
  strongest_unknown: problem | audience | value | concept-scope | game-pillars | narrative | unknown
  downstream_target: bmad | task-planning | marketing-automation | bmad-gds | deck | unknown
  constraint: time | scope | evidence | differentiation | alignment | unknown

If the packet is incomplete, continue with explicit assumptions instead of stalling.

Step 2: Choose one primary mode

Select the mode that reduces ambiguity fastest.

Do not mix multiple modes unless the user explicitly asks for a two-pass workflow.

Step 3: Produce one concept artifact

Return exactly one primary artifact:

• problem brief
• positioning brief
• concept brief
• game concept packet
• story / pitch packet

The artifact should be concise, reusable, and ready to hand to a downstream skill.

Step 4: Make route-outs explicit

When the concept artifact is done, recommend the next downstream lane if appropriate:

• bmad → when the next job is phase routing and selecting PRD / tech spec / architecture / sprint-plan depth
• task-planning → when the concept is approved and needs backlog slices, milestones, or execution packets
• marketing-automation → when the concept is sound and now needs launch, messaging operations, campaign structure, or KPI-aware marketing follow-through
• bmad-gds → when the game concept is strong enough to turn into milestones, GDD slices, playtest planning, or production coordination

Always state why that route-out is next and what packet should be passed forward.

Step 5: Use this output structure

# Idea Routing Brief

## Scope
- Domain: ...
- Stage: ...
- Strongest unknown: ...
- Downstream target: ...
- Confidence: high | medium | low

## Chosen mode
- problem-framing | audience-and-value-framing | concept-shaping | game-concept-framing | story-packaging

## What matters most now
- 2-4 bullets grounded in the packet

## Recommended artifact
- problem brief | positioning brief | concept brief | game concept packet | story / pitch packet

## Artifact contents
| Section | Decision | Why it matters now |
|---------|----------|--------------------|
| ... | ... | ... |

## Immediate next steps
1. ...
2. ...
3. ...

## Route-outs
- Skill / workflow: ...
- Why next: ...
- What to pass forward: ...

## What not to do yet
- 1-3 bullets preventing premature planning, launch detail, or execution drift

Step 6: Keep the artifact lean and decision-oriented

Good outputs are short enough to reuse.

Guardrails:

• prefer a few strong decisions over a giant brainstorm transcript
• convert ideas into constraints, assumptions, and next artifacts
• preserve uncertainty explicitly instead of pretending the concept is resolved
• stop before roadmap depth, sprint decomposition, or campaign execution unless the user asks for the handoff skill next

Compatibility commands and legacy naming

Older BMAD-CIS usage may mention these commands or personas:

• bmad-cis-brainstorming
• bmad-cis-design-thinking
• bmad-cis-innovation-strategy
• bmad-cis-problem-solving
• bmad-cis-storytelling
• Carson / Maya / Victor / Dr. Quinn / Sophia

Treat them as mode hints, not separate first-class workflows. Map them into the modern modes above:

• brainstorming / design thinking / problem solving → usually problem framing or concept shaping
• innovation strategy → usually audience and value framing or concept shaping
• storytelling → usually story packaging

Output format

Always return a short operator-style idea routing brief.

Required qualities:

• one chosen mode, not a kitchen sink of creativity tactics
• one reusable artifact, not a pile of disconnected ideas
• explicit route-outs to neighboring skills when the next lane is clear
• visible assumptions and unresolved questions
• keep the result under roughly 400-700 words unless the user asks for a larger concept packet

Examples

Example 1: product concept before planning

Input

Use bmad-idea. We think there is an opportunity around AI-assisted team handoffs, but we do not know the sharp problem or who it is really for.

Output sketch

• Chosen mode: problem framing
• Recommended artifact: problem brief
• Route-out after approval: bmad

Example 2: GTM / positioning ambiguity

Input

We built a workflow tool, but our messaging is fuzzy and every pitch sounds generic.

Output sketch

• Chosen mode: audience and value framing
• Recommended artifact: positioning brief
• Route-out after approval: marketing-automation

Example 3: game concept shaping

Input

Help us shape a cozy automation game idea before we start a real GDD.

Output sketch

• Chosen mode: game concept framing
• Recommended artifact: game concept packet
• Route-out after approval: bmad-gds

Best practices

1. Act like a concept router, not a prompt zoo — choose one framing mode that fits the ambiguity.
2. Prefer one artifact that travels well — the point is to make downstream planning easier.
3. Use neighboring skills on purpose — bmad, task-planning, marketing-automation, and bmad-gds each own later-stage work.
4. Keep uncertainty visible — a concept brief should surface assumptions, not bury them.
5. Bridge domains without flattening them — product, GTM, and game ideas share structure, but their outputs differ.
6. Treat legacy commands as compatibility language — the modern source of truth is the routing model.

References

• references/operating-modes.md
• references/handoff-boundaries.md
• references/concept-packet-template.md
• ../bmad/SKILL.md
• ../task-planning/SKILL.md
• ../marketing-automation/SKILL.md
• ../bmad-gds/SKILL.md