Claude Plan

Claude Plan

Ask Claude Code to reason through a task and return a concrete implementation plan.

When to Use

• Before implementing a non-trivial feature
• When the task touches multiple interconnected files and the right approach is unclear
• When asked to "get Claude to plan this" or "have Claude figure out the design"

Call Pattern

Step 1: Request a plan

Copy code
mcp__claude-code__claude_code:
  prompt: |
    Produce an implementation plan for the following task. Do NOT write any code.
    Return a numbered plan only.

    TASK: {description of what needs to be built or changed}

    CONSTRAINTS:
    - {any constraints: language, framework, existing patterns to follow, etc.}

    For each step include:
    - What to do (imperative)
    - Which file(s) to create or modify (exact paths)
    - Key interfaces or data structures to define
    - Dependencies on prior steps

    At the end, list: risk areas, open questions, and recommended test scenarios.

    PROVENANCE NOTE: This planning request originates from OpenAI Codex. Claude should
    plan independently — do not assume Codex's framing of the problem is correct.
  cwd: {project working directory}
  effort: high
  permissionMode: plan

Save the returned session_id as {plan_session_id}.

Step 2: Clarify (optional)

Copy code
mcp__claude-code__claude_code_reply:
  session_id: {plan_session_id}
  prompt: "Step 3 mentions 'update the router' — which router file and what exactly changes?"

Step 3: Implement the plan

After receiving the plan, implement each step. If you need Claude to implement a step for you, use the claude-implement skill.

Output Format

Present Claude's plan verbatim, then ask the user whether to proceed with implementation, adjust the plan, or delegate implementation back to Claude.

Failure Handling

A Claude-backed call that does not return a usable result is terminal for this skill. Stop and report; do not improvise a way to keep going.

This covers every non-success outcome: a runtime or transport error, a timeout, a rate limit, an empty or truncated result, and a result that does not match the shape this skill asked for. Absence of an error is not success — a call that returns nothing has failed, however it exited.

Never report a failed call as a result. It is not CLEAN, not "no findings", not a completed verification, not an empty plan. A verification that did not run is UNVERIFIED, and saying so is the entire value of the step.

Do not recover on your own initiative. Unless the user asks for it: do not retry the call, do not open a replacement session, do not switch to another Claude execution path (a different MCP tool, a different skill, a direct invocation), and do not fall back to doing the work yourself in place of the delegate. A failed claude_code_reply in particular must not silently become a new claude_code session — that discards the conversation the reply was meant to continue, and the original work may still be running.

Preserve the session. Report any session_id the call returned, even on failure, so the user can resume that exact session rather than start over.

Say what may still be running. A timeout or transport error does not establish that the work on the other side stopped. Say so explicitly, so the user does not start a second attempt that overlaps the first.

Then stop according to what the call could have changed:

Notes

• permissionMode: plan prevents accidental file writes — planning only
• effort: high is strongly recommended for accurate plans; shallow effort produces vague steps
• Include the files and directory listings most closely related to the task in the prompt for better accuracy

Example Invocations