{
  "skill": {
    "id": "docx-pdf-extract",
    "name": "docx-pdf-extract",
    "description": "Use when you need to extract, compare, or summarize content from PDF and DOCX files with an evidence-first workflow.",
    "path": "skills/research/docx-pdf-extract/SKILL.md",
    "tags": [
      "pdf",
      "docx",
      "extraction",
      "evidence",
      "research"
    ]
  },
  "document": {
    "path": "skills/research/docx-pdf-extract/SKILL.md",
    "frontmatter": {
      "name": "docx-pdf-extract",
      "description": "Use when you need to extract, compare, or summarize content from PDF and DOCX files with an evidence-first workflow.",
      "version": "0.1.0",
      "author": "Hermes Agent",
      "license": "private",
      "metadata": {
        "hermes": {
          "tags": [
            "pdf",
            "docx",
            "extraction",
            "evidence",
            "research"
          ],
          "related_skills": [
            "internal-deep-research",
            "feishu-doc-reporting"
          ]
        }
      }
    },
    "content": "---\nname: docx-pdf-extract\ndescription: Use when you need to extract, compare, or summarize content from PDF and DOCX files with an evidence-first workflow.\nversion: 0.1.0\nauthor: Hermes Agent\nlicense: private\nmetadata:\n  hermes:\n    tags: [pdf, docx, extraction, evidence, research]\n    related_skills: [internal-deep-research, feishu-doc-reporting]\n---\n\n# DOCX + PDF Extract\n\n## Overview\n\nUse this skill to pull structured information from PDF and DOCX documents, especially when the goal is to support research, reporting, or evidence comparison.\n\n## Workflow\n\n1. Identify the exact question.\n2. Read the document structure first if available.\n3. Extract headings, tables, dates, names, and numeric claims.\n4. Normalize text into notes.\n5. Compare multiple files when needed.\n6. Summarize only after evidence has been collected.\n\n## Best Practices\n\n- Prefer source text over OCR when possible\n- Preserve page numbers or section references\n- Separate extracted text from interpretation\n- Flag ambiguous or low-confidence passages\n\n## Verification\n\n- [ ] File type identified\n- [ ] Key sections extracted\n- [ ] Important numbers and names preserved\n- [ ] Summary tied back to source locations\n"
  }
}