Update to Compound Writing

{
  "description": "Compose existing Compound Writing skills and runtime capabilities for open-ended writing outcomes that no single skill covers, such as cross-draft analysis, source-to-outline comparison, or multi-pass revision.",
  "included_files": [],
  "name": "cw-emergent",
  "skill_md_contents": "---\nname: cw-emergent\ndescription: Compose existing Compound Writing skills and runtime capabilities for open-ended writing outcomes that no single skill covers, such as cross-draft analysis, source-to-outline comparison, or multi-pass revision.\n---\n\n# Emergent Capability\n\n## Purpose\n\nEnable the agent to handle requests that don't map directly to a specific skill by composing primitives and existing skills into novel solutions.\n\nThis is what makes the system agent-native: the ability to figure out how to accomplish an outcome, not just execute predefined workflows.\n\nRead `../../references/context-contract.md` before composing a multi-artifact workflow. Respect the active project's source-of-truth files, provenance, destinations, and approval rules.\n\n## When This Applies\n\nUse emergent composition when:\n- Request spans multiple drafts or files\n- Request requires pattern analysis across the body of work\n- Request combines multiple types of feedback\n- No single skill handles the full request\n- User describes an outcome, not a command\n\nDo NOT use when:\n- A skill directly handles the request (use the skill)\n- The request is a simple file operation\n- The user invokes a specific command\n\n## The Composition Loop\n\nWhen a request doesn't map to a skill:\n\n1. **Understand the outcome**\n   - What does success look like?\n   - What would the user see/know when this is done?\n\n2. **Identify capabilities needed**\n   - What artifacts must be found, read, compared, or revised?\n   - What maintained context or connected source must be loaded?\n   - What draft, source, or destination operations are authorized?\n   - What analysis operations? (compare, analyze voice, check AI)\n\n3. **Identify skills that help**\n   - Which existing skills apply to parts of the request?\n   - Can skills be chained for the full solution?\n\n4. **Compose a solution**\n   - Chain primitives and skills together\n   - Plan the sequence of operations\n\n5. **Execute in a loop**\n   - Work through the plan\n   - Adapt if something unexpected comes up\n   - Continue until outcome achieved or blocked\n\n6. **Report results**\n   - Summarize what was found/done\n   - Offer next steps if applicable\n\n## Example Requests and Approaches\n\n| Request | Primitives | Skills | Approach |\n|---------|-----------|--------|----------|\n| \"Compare voice in my last 3 drafts\" | Find and read the relevant pieces | `cw-voice-check` | Analyze each against the same live voice sources, then compare |\n| \"Find patterns in my hooks\" | Find and read the relevant pieces | — | Extract openings, classify moves, and report with examples |\n| \"What objections keep coming up?\" | Read the relevant drafts and reviews | `cw-objections` | Compare repeated resistance and distinguish writer pattern from topic pattern |\n| \"Am I overusing certain transitions?\" | Search the requested corpus | `cw-transition` | Count, inspect context, and report only meaningful repetition |\n| \"Make this article better\" | Load the draft, brief, voice, and sources | `cw-dev-edit`, `cw-line-edit`, `cw-ai-check` | Diagnose, revise, then validate |\n| \"Help me finish this piece\" | Inspect the active artifact and open decisions | varies | Route to the smallest unfinished outcome |\n| \"Cross-reference notes with outline\" | Read both authoritative artifacts | — | Compare claims, examples, sources, and omissions |\n\n## Judgment: When to Compose\n\n**Good candidates for composition:**\n- \"Compare X across Y\" — Analysis pattern\n- \"Find patterns in Z\" — Discovery pattern\n- \"Check if A matches B\" — Validation pattern\n- \"Make this better\" — Improvement pattern\n- \"Help me with X\" (vague) — Needs investigation first\n\n**Not good candidates:**\n- \"Give me a hook\" → Use /cw-hook directly\n- \"Does this bury the lede?\" → Use /cw-bluf directly\n- \"Check my voice\" → Use /cw-voice-check directly\n- \"Run the nemesis read\" → Use /cw-nemesis directly\n\n## The Ultimate Test\n\nCan the agent accomplish tasks you didn't explicitly design for?\n\n**Example:** User asks \"Cross-reference my meeting notes with my task list and tell me what I've committed to but haven't scheduled.\"\n\nThis wasn't designed as a feature. But if the agent can:\n1. Read the meeting notes\n2. Read the task list\n3. Extract commitments from notes\n4. Compare against tasks\n5. Report the gaps\n\n...then it's agent-native.\n\n## Failure Modes\n\nIf composition fails:\n- **Missing capability or access:** Report what is unavailable\n- **Skill not available:** Suggest manual alternative\n- **Too complex:** Break into smaller requests\n- **Ambiguous:** Ask for clarification\n\nNever say \"I don't have a feature for that\" when composition could work.\n\n## Reporting Results\n\nWhen composition completes, report:\n1. What was requested\n2. What was done (briefly)\n3. What was found/achieved\n4. Suggested next steps (if applicable)\n\nKeep reports focused on insights, not process details.\n\n## Lessons\n\n[Skill-specific lessons will be added here as they're captured]\n"
}