What problem does it solve? Initial implementation plans often miss tasks, contain vague descriptions, or have wrong dependency ordering. This Skill performs a second-pass review of an existing plan against the actual codebase, surfacing gaps, redundancies, and out-of-scope work before implementation begins. ## Core Features & Use Cases - Deep Plan Audit: Re-analyzes the codebase to find missing tasks (migrations, config updates, edge cases), vague task descriptions, wrong dependencies, and duplicate or out-of-scope tasks. - Optional Validation Pass: The +check flag dispatches a fresh-context subagent that verifies each finding against the repository, dropping invented items and correcting inaccurate ones. - Plan Discovery Modes: Use --list for read-only plan discovery, @<path> to target a specific plan file, or automatic branch-based resolution with slug or sequential naming. - Use Case: After generating a feature plan with /aif-plan, run /aif-improve to catch that the plan forgot a database migration task and that Task #5 should depend on Task #2, then apply the approved fixes directly to the plan file. ## Quick Start Ask the AI to run /aif-improve on the current plan to find missing tasks and fix dependencies before implementation.
