What problem does it solve? Writing implementation plans without understanding the actual codebase leads to vague specs, missed edge cases, and scope creep. This Skill guides an AI agent through a structured, interactive process that grounds every plan in verified code research before writing a single line of the specification. ## Core Features & Use Cases - Codebase-Grounded Research: Spawns parallel research agents (codebase-locator, codebase-analyzer, thoughts-locator) to find relevant files, trace data flow, and return file:line references before planning begins. - Iterative Plan Development: Builds plans in stages—context gathering, research, outline approval, detailed writing—with user checkpoints at each step to allow course corrections. - Structured Plan Output: Writes plans to thoughts/shared/plans/ with a strict template covering current state, phased changes, automated vs. manual success criteria, and explicit out-of-scope items. - Use Case: Given a ticket like thoughts/allison/tickets/eng_1478.md describing parent-child tracking for sub-tasks, the Skill reads the ticket fully, researches the daemon codebase, presents design options, and produces a phased implementation plan with verifiable success criteria. ## Quick Start Ask the agent to create an implementation plan for a ticket file, for example: run ff_create_plan on thoughts/allison/tickets/eng_1234.md and iterate with me on the design.
