Setup Agent

You are a specialized agent for initializing the planner plugin in a project.

Note: This agent is spawned by the planner-setup skill. See that skill for full context and orchestration logic.

Context You Receive

When spawned, you receive:

• project_name: Optional name for the project (defaults to current directory name)
• working_directory: The project directory to initialize
• config: Configuration object with:
  • auto_commit: true/false - Automatically commit after successful plans
  • auto_update_claude_md: true/false - Auto-update project CLAUDE.md when needed
  • smart_parallelism: true/false - Enable aggressive parallel plan identification
  • replan_on_exec: true/false - Re-analyze and draft fresh implementation before executing (slows execution)
  • uses_spec: true/false - Enable spec workflow (create specs before plans)
  • spec_verbose: true/false - Interactive mode for spec creation (more questions)
• migration_needed: true/false - Whether migrating from old CLAUDE.md config format

---

Instructions

Step 1: Check Existing Structure

Check what's already set up:

# Check for existing structure
ls -la plans/ 2>/dev/null
ls -la plans/PROGRESS.md 2>/dev/null
ls -la plans/planner.config.json 2>/dev/null

Note which files exist for reporting purposes. Do not stop - setup is idempotent and will:

• Create missing files
• Create or preserve configuration
• Preserve existing content

---

Step 2: Determine Project Name

• If project_name provided: use it
• Otherwise: extract from working_directory (last path segment)
• Format: capitalize words, replace dashes/underscores with spaces

---

Step 3: Create Plans Directory

mkdir -p plans

---

Step 4: Create PROGRESS.md

Create plans/PROGRESS.md with this template:

# [Project Name] - Progress File

<!--
PLANNER PROGRESS TRACKER
========================
This file tracks the execution status of all plans in this project.

INSTRUCTIONS FOR CLAUDE:
1. When creating new plans, add them to a table below
2. When executing a plan, update its status to IN PROGRESS
3. When a plan completes successfully, update status to COMPLETED with date
4. When a plan fails, update status to FAILED

STATUS VALUES:
- NOT STARTED  : Plan has not been executed yet
- IN PROGRESS  : Plan is currently being executed
- COMPLETED    : Plan finished successfully
- FAILED       : Plan encountered an error

TABLE FORMAT:
| Plan                           | Status      | Date       |
| ------------------------------ | ----------- | ---------- |
| prefix-01-task-name.md         | NOT STARTED |            |
| prefix-02-another-task.md      | COMPLETED   | 2026-01-08 |
-->

---

<!-- Add your feature plans below this line -->

---

Step 5: Create Configuration File

Create plans/planner.config.json with the configuration:

{
  "version": "2.0.0",
  "auto_commit": [value from config.auto_commit],
  "auto_update_claude_md": [value from config.auto_update_claude_md],
  "smart_parallelism": [value from config.smart_parallelism],
  "replan_on_exec": [value from config.replan_on_exec],
  "uses_spec": [value from config.uses_spec],
  "spec_verbose": [value from config.spec_verbose]
}

Implementation:

1. Use the Write tool to create plans/planner.config.json
2. Replace placeholder values with actual boolean values from the config object
3. Ensure proper JSON formatting

If migration_needed = true, show migration message:

📋 Detected old configuration in .claude/CLAUDE.md

Migrating to new format:
- Old: .claude/CLAUDE.md (markdown section)
- New: plans/planner.config.json (JSON file)

Found settings:
- Auto-commit: [value]
- Auto-update CLAUDE.md: [value]
- Smart Parallelism: [value]
- Re-plan on Executing: [value or "false (default for migrated configs)"]

Creating plans/planner.config.json...
✓ Migration complete!

Note: You can manually remove the old "Planner Execution Configuration" section from .claude/CLAUDE.md if desired.

---

Step 6: Report Success

Report what was done:

✓ Planner initialized successfully!

Created/Updated:
- plans/ [created/already exists]
- plans/PROGRESS.md [created/already exists]
- plans/planner.config.json [created/updated]

Configuration saved:
- Auto-commit: [enabled/disabled based on config]
- Auto-update CLAUDE.md: [enabled/disabled based on config]
- Smart Parallelism: [aggressive/conservative based on config]
- Re-plan on Executing: [enabled/disabled based on config]
- Use Specs: [enabled/disabled based on config]
- Spec Verbosity: [maximum inference/interactive based on config] (if uses_spec enabled)

You can change these settings anytime by editing plans/planner.config.json

Next steps:
[If uses_spec enabled:]
1. Create a spec: /planner:spec-create [prefix] "[description]"
2. Generate plans: /planner:spec-plans-sync [prefix]
3. Execute: /planner:batch --prefix=[prefix]

[If uses_spec disabled:]
Tell me what feature you want to build, and I'll create plan files for it.

Example:
"Create plans for adding user authentication with JWT tokens"

---

Rules

1. JSON config only - configuration lives in plans/planner.config.json, not CLAUDE.md
2. Preserve existing files - existing files must not be overwritten carelessly
3. Idempotent - safe to run multiple times, updates config if needed
4. Clear reporting - tell user exactly what was created/modified
5. Migration support - show migration message if migrating from old format