AgentSkillsCN

bmad-discovery-research

头脑风暴想法并研究项目。

SKILL.md
--- frontmatter
name: bmad-discovery-research
description: Brainstorms ideas and researches projects.
allowed-tools: ["Read", "Write", "Grep"]
metadata:
  auto-invoke: true
  triggers:
    patterns:
      - "I have an idea"
      - "What if we"
      - "Help me think"
      - "explore possibilities"
      - "I'm thinking about"
      - "brainstorm"
      - "research"
    keywords:
      - idea
      - brainstorm
      - explore
      - research
      - thinking
      - discovery
      - analyze
  capabilities:
    - brainstorming
    - research
    - competitive-analysis
    - discovery
    - problem-exploration
  prerequisites: []
  outputs:
    - discovery-brief
    - research-notes
    - problem-statement

Discovery Analysis Skill

When to Invoke

Automatically activate this skill when the user:

  • •Says "I have an idea...", "What if we...", "I'm thinking about..."
  • •Asks "Help me think through...", "Can you help me brainstorm...", "Let's explore..."
  • •Mentions "new project", "new feature", "explore possibilities"
  • •Talks about research, competitive analysis, or market exploration
  • •Has vague requirements that need clarification
  • •Mentions understanding or documenting an existing project
  • •Is starting a Level 3-4 project (complex/novel problems)
  • •Uses words like: idea, brainstorm, explore, research, discovery, analyze

Specific trigger phrases:

  • •"I have an idea for [something]"
  • •"What if we built [something]"
  • •"Help me think through [problem]"
  • •"Can we explore [opportunity]"
  • •"I need to research [topic]"
  • •"Document this project"
  • •"Understand this codebase"

Do NOT invoke when:

  • •User already has a detailed PRD (use bmad-product-planning instead)
  • •User is asking for implementation help (use bmad-development-execution instead)
  • •User has clear, well-defined requirements (skip to bmad-product-planning)
  • •User is asking about workflow status (use main-workflow-router)
  • •Project is Level 0-2 and requirements are clear (skip to bmad-product-planning)

Mission

Turn vague ideas or problem statements into structured briefs that downstream skills can trust. Identify goals, constraints, risks, and unknowns to inform product planning.

Inputs Required

  • •problem_statement: initial idea, pain, or opportunity description
  • •stakeholders: who cares about the outcome and why
  • •context_assets: repos, documents, or market references available for analysis

If essential context is missing, gather it before deeper synthesis.

Outputs

  • •Discovery brief following patterns in REFERENCE.md
  • •Brainstorm notes (from assets/brainstorm-template.md.template)
  • •Product brief (from assets/product-brief-template.md.template)
  • •Research dossier (from assets/research-dossier-template.md.template)
  • •Prioritized questions and risk register captured for product-requirements skill
  • •Recommendation on readiness to progress or need for further discovery

Template locations: .claude/skills/bmad-discovery-research/assets/*.template

Process

  1. •Validate entry criteria in CHECKLIST.md and classify project complexity.
  2. •Conduct desk research across provided assets; cite sources.
  3. •Frame insights into concise problem summary, goals, personas, and constraints (use templates in assets/).
  4. •Document open questions, assumptions, and risks with recommended owners.
  5. •Deliver summary plus links to created artifacts for the orchestrator and stakeholders.

Note on automation: This skill currently operates through collaborative conversation using templates. No automation scripts are required—artifacts are generated manually using Jinja2 templates from assets/. See scripts/README.md for future automation roadmap.

Quality Gates

Confirm CHECKLIST.md is satisfied before signaling readiness for planning. Missing data or low confidence requires escalation.

Error Handling

  • •If the idea remains ambiguous after initial probing, request specific clarifications rather than guessing.
  • •Flag conflicting stakeholder goals and recommend alignment conversations.
  • •When scope is too small (Level 0-1), suggest redirecting to lighter-weight workflows documented in REFERENCE.md.