What problem does it solve? It turns ad-hoc chat messages like "task:" or "fix:" into a persistent, queryable task database, then executes those tasks through isolated background agents with validation, repair, and restart safeguards instead of losing track of work in conversation history. ## Core Features & Use Cases - Task capture and tracking: Log tasks, fixes, and todos with priorities, tags, requirements, and dependencies via the task-db helper, then list, check, complete, cancel, or reprioritize them. - Agent-based execution: Dispatch read-only Planning Scouts and write-capable Executors in isolated worktrees, with ownership tracking, snapshot validation, and a strict accept/validate/reject decision flow. - Plan management and changelogs: Group tasks under plan: epics, reconcile plan documents with tasks through a propose/apply loop, and auto-generate CHANGELOG.md entries from completed work. - Use Case: Say "fix: login button crashes on empty password" and the skill logs it as a high-priority fix, offers to run it immediately, dispatches a Scout and Executor in a worktree, validates the result, and commits only task-owned changes after your confirmation. ## Quick Start Tell the assistant "task: add dark mode toggle to settings page" and choose Run Now when prompted to have it logged, planned, executed, and validated.
