Paths: File paths (references/, ../ln-*) are relative to this skill directory.

Epic Coordinator

Type: L2 Domain Coordinator Category: 2XX Planning

Universal Epic management coordinator that handles both creation and replanning through scope decomposition.

Purpose

Coordinates Epic creation (CREATE) and replanning (REPLAN) from scope decomposition. Discovery and research stay read-only until preview is approved; execution remains inline inside ln-210.

When to Use This Skill

This skill should be used when:

• Start new scope/initiative requiring decomposition into multiple logical domains (CREATE mode)
• Break down large architectural requirement into Epics
• Update existing Epics when scope/requirements change (REPLAN mode)
• Rebalance Epic scopes within an initiative
• Add new Epics to existing initiative structure
• First step in project planning (scope → Epics → Stories → Tasks)
• Define clear scope boundaries and success criteria for each domain

Output: 3-7 tracker Epics (logical domains/modules; transport per the configured provider)

Runtime Contract

MANDATORY READ: Load references/coordinator_runtime_contract.md, references/epic_planning_runtime_contract.md, references/epic_plan_summary_contract.md MANDATORY READ: Load references/researchgraph_mcp_usage.md when the project has docs/hypotheses/, docs/goals/, or benchmark run manifests that can change Epic boundaries.

Runtime family: epic-planning-runtime

Identifier:

• scope identifier

Phases:

1. PHASE_0_CONFIG
2. PHASE_1_DISCOVERY
3. PHASE_2_RESEARCH
4. PHASE_3_PLAN
5. PHASE_4_MODE_DETECTION
6. PHASE_5_PREVIEW
7. PHASE_6_DELEGATE
8. PHASE_7_FINALIZE
9. PHASE_8_SELF_CHECK

Decision handling:

• preview/confirmation is runtime PAUSED + pending_decision
• ln-210 writes final epic-plan summary for ln-200 when used as downstream coordinator

Coordinator artifact flow:

• execution remains inline inside ln-210
• PHASE_7_FINALIZE writes the coordinator summary through node references/scripts/epic-planning-runtime/cli.mjs record-plan-summary
• managed parent flow stores the artifact at .hex-skills/runtime-artifacts/runs/{parent_run_id}/epic-plan/{identifier}.json

Worker Invocation (MANDATORY)

Downstream contract: ln-210 is itself a worker for ln-200 and writes epic-plan machine-readable output after inline Epic create/replan execution completes.

TodoWrite format (mandatory)

- Phase 1: Discover scope and project context (pending)
- Phase 2: Research only what changes Epic boundaries (pending)
- Phase 3: Build ideal Epic plan (pending)
- Phase 4: Detect mode and validate Infra Epic need (pending)
- Phase 5: Preview and confirm plan (pending)
- Phase 6: Execute create/replan flow (pending)
- Phase 7: Finalize epic-plan summary (pending)
- Phase 8: Self-check (pending)

Core Concepts

Decompose-First Pattern

Key principle: ALWAYS analyze scope and build IDEAL Epic plan FIRST, THEN check existing Epics to determine mode:

• No existing Epics → CREATE MODE (generate and create all Epics)
• Has existing Epics → REPLAN MODE (compare, determine operations: KEEP/UPDATE/OBSOLETE/CREATE)

Rationale: Ensures consistent Epic decomposition based on current scope requirements, independent of existing Epic structure (which may be outdated or suboptimal).

Epic 0 Reserved for Infrastructure

MANDATORY READ: Load references/numbering_conventions.md for Epic 0 rules, when to use it, and tracker numbering.

Inputs

Resolution: Epic Resolution Chain (adapted: scope doc discovery). Fallback: IF no scope doc found → AskUserQuestion: "What should be decomposed into Epics?"

Workflow

Phase 0: Tools Config

MANDATORY READ: Load references/environment_state_contract.md, references/storage_mode_detection.md, references/input_resolution_pattern.md

Extract: task_provider = Task Management → Provider

Phase 1: Discovery

Objective: Gather the minimum context required to shape Epic boundaries before any preview or mutation.

Step 0: Resolve scopeDoc (per input_resolution_pattern.md, adapted for scope):

• IF args provided (scope description or doc path) → use args
• ELSE IF docs/project/requirements.md exists → use as scope source
• ELSE IF docs/requirements.md exists → use as scope source
• ELSE → AskUserQuestion: "What should be decomposed into Epics?"

Step 1: Load Configuration

Auto-discovers Team ID and Next Epic Number from docs/tasks/kanban_board.md:

• Team ID: Reads Tracker Configuration table → Fallback: Ask user directly
• Next Epic Number: Reads Next Epic Number field → Fallback: Ask user directly

MANDATORY READ: Load CLAUDE.md — sections "Configuration Auto-Discovery" and "Tracker Integration".

Phase 2: Research

Objective: Research only what changes Epic boundaries, Infra Epic need, or batch preview quality.

Step 1: Project Research

Objective: Research project documentation AND frontend code to understand context BEFORE asking user questions.

Process:

0. Researchgraph Preflight (if present):

   • If docs/hypotheses/, docs/goals/, or benchmark/runs/*/manifest.yaml exists, run read-only verify_index first.
   • Use inspect_goal, trace_goal_tree, find_hypotheses, or audit_goal_alignment only when goals/hypotheses can change Epic boundaries, validation scope, or promotion/generalization work.
   • Treat STALE graph debt as planning context, not as an Epic blocker by itself.

1. Document Scan:

   • Use Glob to find: docs/requirements.md, docs/architecture.md, docs/tech_stack.md
   • Use Read to load found documents

2. Frontend Code Scan (if applicable):

   • Use Glob to find: **/*.html, src/**/*.html, public/**/*.html, templates/**/*.html
   • Use Read to load HTML files
   • Extract functional domains from:
     • Navigation menus: <nav>, <a href> links reveal feature areas
     • Forms: Input fields reveal data models (user registration, login, checkout)
     • Page titles: <h1>, <title> tags reveal feature names
     • Route patterns: URL structures reveal domain boundaries

   Example HTML extraction:

   <nav>
     <a href="/products">Products</a>
     <a href="/cart">Shopping Cart</a>
     <a href="/checkout">Checkout</a>
   </nav>
   <!-- Reveals domains: Product Catalog, Shopping Cart, Payment -->
   

3. Extract key information from docs + HTML:

   • Business objectives: What is the project trying to achieve? (from requirements.md)
   • User personas: Who will use the system? (from requirements.md)
   • Major functional domains: What are the main modules/areas? (from requirements.md, architecture.md, HTML navigation)
   • Technical stack: What technologies mentioned? (from tech_stack.md, architecture.md, HTML meta/script tags)
   • Infrastructure requirements: Any mention of logging, monitoring, deployment, CI/CD, security, performance optimization?

4. Combine findings:

   • Merge domains from docs + HTML (deduplicate, consolidate similar)
   • Example: "User Auth" (from docs) + "Login" (from HTML) → "User Management"

Fallback: If docs AND HTML missing → Skip to scope questioning; do not invent research-heavy detail

Step 2: Infrastructure Epic Decision

Objective: Determine if Infrastructure Epic (Epic 0) should be proposed.

Criteria for Infrastructure Epic:

✅ PROPOSE Infrastructure Epic (Epic 0) if ANY of:

1. New project (no docs/infrastructure.md found, no Epic "Infrastructure" in ka