1. Parse the user request to identify the plan file and task subset.
2. Read and parse the plan file, extracting task details and context.
3. Build a context pack for each task, including file paths, naming conventions, and dependencies.
4. Launch Sparky subagents in a rolling pool (max 12), assigning tasks with context packs.
5. Validate the output of each subagent and update the plan file.
6. Perform a final integration pass to reconcile conflicts and ensure codebase integrity.
7. Run tests and fix failures until all tests pass.
8. Provide a completion summary with task status, issues, and resolutions.