What problem does it solve? Implementation work often stalls mid-run because plans skip decisions, ignore git history shape, or force the executing agent to stop and ask questions. This Skill produces a plan file where every decision is front-loaded, each step maps to an atomic commit, and execution can run unattended under a persistent implementation subagent. ## Core Features & Use Cases - Decision front-loading: Surfaces every user-facing naming, edge-case, and scope decision via AskUserQuestion before entering plan mode, so execution never stalls. - Commit-shaped planning: Loads the git-workflow skill and structures each plan step as one intended atomic commit with verification steps and commit messages. - Check tiers and validation: Assigns each step a direct or validation-subagent check tier, names the implementation model (opus or fable), and ends with a final whole-branch validation checkpoint plus MR steps. - Use Case: After a pre-plan briefing and design discussion, run the command with a topic or ticket ref to produce a plan file that an orchestrator and implementation subagent execute step-by-step through MR open and babysit. ## Quick Start Ask Claude to write an implementation plan for your feature or ticket, for example: write a plan for adding rate limiting to the API based on our pre-plan briefing.
