research

User Input

$ARGUMENTS

You MUST consider the user input before proceeding.

Outline

Phase 0: Setup

1. Parse topic from $ARGUMENTS

   • If empty, ask user what to research

2. Auto-detect research mode based on topic keywords:

   • Competitor Analysis: "alternatives", "vs", "competitors", "pricing", "compare", product/company names
   • Technical Research: "API", "how to", "best practices", "parsing", "library", "framework", "implementation", "github"
   • If unclear, proceed with general research approach

3. Generate output path:

   • Slugify topic: lowercase, replace spaces with -, remove special chars
   • Path: {REPO_ROOT}/research/{topic-slug}.md
   • Create directory if not exists: mkdir -p research

4. Initialize research document with metadata (atomic write):

   # Research: {Topic}
   
   **Date**: {YYYY-MM-DD}
   **Mode**: {Detected Mode}
   **Status**: In Progress
   

5. Generate subtopics (query decomposition):

   • Decompose main topic into 3-5 focused subtopics
   • Adapt subtopics based on detected mode:

   For Competitor Analysis:

   • What is {topic}? (market category, core value prop)
   • Who are the main players? (direct competitors)
   • How do they compare? (features, pricing, positioning)
   • What are the problems? (complaints, limitations)
   • What do users recommend? (community preferences)

   For Technical Research:

   • What is {topic}? (definition, core concepts)
   • How does it work? (mechanics, implementation)
   • What are the alternatives? (other approaches)
   • What are the problems? (limitations, gotchas)
   • What are best practices? (recommendations, patterns)

   Show the research plan to user before starting deep research.

Phase 1: Scope Clarification (if needed)

Only ask clarifying questions if topic is genuinely ambiguous. Max 3 questions.

Present questions one at a time with recommendation:

Recommended: Option X - {reasoning}