name: mip-annotate description: Phase 2 of MIP pipeline - Orchestrate parallel annotation of all 9 chapters using subagents. Each chapter analyzed independently for metaphors. allowed-tools: Task, Read, Write, Bash

MIP Annotation Phase

Orchestrate parallel metaphor annotation across all 9 chapters of The Great Gatsby.

Your Task

Spawn 9 parallel subagents using the Task tool to analyze each chapter independently. Each subagent applies the MIP (Metaphor Identification Procedure) and outputs structured JSON.

Critical Implementation

You MUST use the Task tool to spawn 9 subagents in parallel. Example:

Use Task tool with subagent_type: "general-purpose" for each chapter

Subagent Instructions Template

Each subagent should receive these instructions:

You are annotating Chapter X of The Great Gatsby for metaphors using MIP.

Read: book/chapters/chapter-0X.txt
Read: mip-analysis/research-notes.md (for procedure and seed systems)

Apply MIP to identify 8-15 significant metaphors. For each:

1. Read surrounding context (full paragraph minimum)
2. Identify lexical unit (word/phrase with potential metaphorical meaning)
3. Determine contextual meaning (meaning in THIS passage)
4. Determine basic meaning (concrete, bodily, historically older sense)
5. Compare: if different and understood via comparison → metaphor
6. Classify MIPVU type: indirect, direct, or implicit
7. Map source → target domains
8. Note thematic function

Output JSON to: mip-analysis/annotations/chapter-0X.json

JSON Schema

Each annotation file must follow this structure:

{
  "chapter": 1,
  "total_metaphors": 12,
  "annotations": [
    {
      "id": "ch1-001",
      "location": {
        "chapter": 1,
        "paragraph": 3,
        "quote": "the exact quoted text"
      },
      "lexical_unit": "the specific word or phrase",
      "contextual_meaning": "what it means in this specific passage",
      "basic_meaning": "concrete, physical, or historically older meaning",
      "mip_decision": true,
      "mipvu_type": "indirect|direct|implicit",
      "source_domain": "SOURCE DOMAIN IN CAPS",
      "target_domain": "TARGET DOMAIN IN CAPS",
      "cmt_mapping": "TARGET IS SOURCE",
      "function": ["theme", "characterization", "setting", "narrative"]
    }
  ]
}

Process

1. Read mip-analysis/research-notes.md to understand the procedure
2. Ensure mip-analysis/annotations/ directory exists
3. Spawn 9 parallel Task agents, one per chapter
4. Each agent reads its chapter and the research notes
5. Each agent outputs its JSON file
6. Verify all 9 files created successfully

Chapters to Process

Success Criteria

• 9 JSON files created in mip-analysis/annotations/
• Each file contains 8-15 metaphor annotations
• Total of 40-80+ metaphors identified across all chapters
• JSON structure valid and consistent
• Ready for /mip-cluster phase