opus-plan

opus-plan

Delegate the planning task to a Claude Opus 4.6 Plan subagent. Uses "ultrathink" in the prompt to maximize reasoning depth. Note: this does NOT use the extended thinking API — effort is elicited via prompt, not configured via budget tokens.

Instructions

1. Identify the task to plan from <command-args> if provided, otherwise from the current conversation context (the most recent user request or problem description).

2. Launch an Agent with these parameters:

   • subagent_type: "Plan"
   • model: "opus"
   • A detailed prompt that includes:
     • The full task description
     • Relevant context from the conversation (files mentioned, constraints, goals)
     • Instruction to be thorough: explore trade-offs, identify risks, consider alternatives, and produce a step-by-step implementation plan
     • Instruction to read relevant files before making recommendations (never plan blind)

3. Generate a timestamp via Bash: date +%Y%m%d-%H%M%S

4. Ensure the ./tmp directory exists: mkdir -p ./tmp

5. Write the full plan to ./tmp/claude-plan-<timestamp>.md using the Write tool. Report the path to the user.

6. Present the plan returned by the subagent to the user. Do not summarize or truncate it.

Example prompt to pass to the subagent

ultrathink

You are planning the implementation of: <task>

Context:
- <any relevant files, constraints, or background from the conversation>

Produce a thorough implementation plan:
- Read all relevant files before recommending changes
- Identify the key files and components involved
- List the steps in order, with specific file paths and line numbers where applicable
- Call out risks, trade-offs, and alternative approaches
- Note any open questions that need the user's input before proceeding