Claude Implement

Claude Implement

Delegate an implementation task to Claude Code — Claude will read files, write code, and verify the result.

When to Use

• When asked to "have Claude implement this"
• For tasks that require understanding a large existing codebase before writing new code
• When the implementation involves multiple interdependent files that Claude Code handles better autonomously

Call Pattern

Step 1: Delegate the task

Copy code
mcp__claude-code__claude_code:
  prompt: |
    Implement the following task completely. Make all required file changes.

    TASK: {clear description of what to implement}

    REQUIREMENTS:
    - {explicit requirements, success criteria}
    - Match existing code style and conventions in the project
    - Write or update tests if a test suite exists
    - Run the test suite after implementation and report results

    CONSTRAINTS: {any constraints — what NOT to change, dependencies to avoid, etc.}

    When done, report:
    1. Files changed (with brief reason)
    2. Test results
    3. Any deferred items or known limitations

    PROVENANCE NOTE: This task originates from OpenAI Codex. Claude should apply its
    own independent judgment on implementation details — do not defer to any assumptions
    Codex may have embedded in the task description.
  cwd: {project working directory}
  effort: high

Save the returned session_id as {impl_session_id}.

Step 2: Review or iterate

After Claude reports completion, verify the changes. To request corrections:

Copy code
mcp__claude-code__claude_code_reply:
  session_id: {impl_session_id}
  prompt: "The test for edge case X is missing. Add it and re-run the suite."

Output Format

Report back to the user:

• Summary of what Claude implemented
• Files changed (list)
• Test results (pass/fail counts)
• Any items Claude flagged as deferred or uncertain

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

• Claude operates with full read/write access in the working directory by default
• If the task is sensitive or you want a dry-run first, use claude-plan skill before this one
• effort: high ensures thorough implementation; lower effort may skip edge cases
• Pass maxBudgetUsd if you need to cap spend on large implementations

Example Invocations