What problem does it solve? Running a spec-driven feature lifecycle manually means invoking each phase skill (/do-brainstorm, /do-design, /do-plan, /do-implement, /do-test, /do-code-review) one at a time and remembering where to stop for approvals. This Skill automates that orchestration: it classifies the task, resolves the correct workflow from the DoFlow runtime, runs each stage in order, and pauses only at declared approval gates. ## Core Features & Use Cases - Runtime-validated task classification: Proposes one of nine task classes (feature, bug, refactor, review, research, documentation, dependency-change, operations, trivial-edit) and validates it through the DoFlow runtime before any stage runs, rejecting mismatched classes instead of defaulting to the feature chain. - End-to-end stage orchestration: Executes the resolved workflow's stages in order, resuming feature work from the first missing artifact (requirement.md, design.md, plan.md) and honoring optional stages and readiness checks. - Approval gate management: Stops at workflow-declared gates (post-discovery clarifications, pre-implementation checkpoint, post-review handoff), records each decision via the orchestrate command, and never proceeds on ambiguous answers. - Use Case: Tell the assistant "build this feature end-to-end" and it will classify the work, run brainstorm through code review automatically, pause for your approval before implementation, and hand off to /do-git for commit once the final gate clears. ## Quick Start Ask the assistant to take your feature idea from idea to implementation using the do-flow skill, for example by saying "run the doflow chain to add OAuth login support".
