What problem does it solve? AI-OS projects track tasks in both Markdown files (TASKS.md, REVIEWS.md) and a SQLite state database, and these two sources can silently drift apart — for example, a task is checked off in TASKS.md but still OPEN in state.sqlite. This Skill detects those mismatches before you trust the task list or claim work is done. ## Core Features & Use Cases - One-step sync verification: Wraps the task-synchronizer-mcp verify_markdown_sync tool and returns a SYNC_PASS or SYNC_FAIL verdict. - Structured anomaly interpretation: Parses the SYNC_RESULT JSON tail (status, anomalies, auto_fixes) and maps each failure pattern to a concrete fix. - Done-claim gate: Blocks "work is done" claims until a SYNC_FAIL is resolved, routing status fixes through the ai-task skill. - Use Case: Before committing, run the check and discover that E-178 is marked [x] in TASKS.md but still OPEN in state.sqlite, then fix it via the ai-task skill instead of shipping inconsistent bookkeeping. ## Quick Start Ask the AI to verify that TASKS.md and the project state database are in sync before marking any task as done.