What problem does it solve? AI coding sessions often skip team conventions: they commit straight to main, skip planning, fabricate code details, and lose task documentation. This Skill enforces a consistent Raw Prompt → Understand → Plan → Code → Document workflow so every change follows the same branch, review, and PR process. ## Core Features & Use Cases - Structured five-phase workflow: Moves a task from raw prompt capture through understanding, execution planning, implementation, and archival, with documents like execution_plan.md and acceptance_criteria.json produced along the way. - Git safety enforcement: Detects attempts to commit on the default branch, blocks bare force-pushes, resolves the default branch dynamically, and auto-pushes task documentation to a separate docs repo with conflict resolution. - Configurable autonomy: Reads .claude/config_hints.json for project namespace, ticket tracker, coding standards location, and opt-in autonomous/continuous modes that gate judgement checkpoints and prompt suppression. - Use Case: A developer says "ar-taskflow" with a task folder path; the assistant reads the project config, drafts an execution plan, creates a feature branch like feature/svc-42-add-auth, implements with scoped verification, and opens a PR against the correct base branch. ## Quick Start Say "ar-taskflow" or provide a task folder path and ask the assistant to run the structured workflow for your task.
