tbroadley/iterate

Iterate: Plan and Implement

Interactive planning session with the user, followed by autonomous implementation. After implementation, the /iterate-review skill handles the review → fix → cleanup → finalize loop.

Planning (Interactive)

Work with the user to create a detailed implementation plan. Tell the user: "I'm going to make this plan very detailed because once I start implementing, I won't ask for any more input."

Planning Requirements

The plan must be detailed enough that no further human input is needed. Push for specifics:

• Exactly which files to create/modify
• Function signatures, data structures, key logic
• How to handle edge cases
• What tests to write
• Acceptance criteria: how to verify the implementation is correct

Ask clarifying questions until the plan is unambiguous.

Plan Output

Enter plan mode and write the finalized plan as a numbered list of concrete implementation steps. Each step should be small enough for a single subagent to execute. Group related changes that must be consistent (e.g., updating a function signature and all its callers) into the same step.

At the end of the plan, add this section:

## Post-Implementation

After all implementation steps are complete and validation passes:
1. Stage and commit all changes with a descriptive message ending with `Co-Authored-By: <agent-name> <agent-noreply-email>`
2. Push to remote
3. Invoke the `/iterate-review` skill to review, clean up, and finalize the changes

This ensures the next agent (after context clearing) knows to invoke the review skill.