What problem does it solve? Running a spec-driven development workflow manually means typing each pipeline command (specify, plan, tasks, implement, mark-complete) and waiting at every review gate, which is slow when you just want a feature spec executed end-to-end without supervision. ## Core Features & Use Cases - Unattended Pipeline Orchestration: Dispatches each per-step /speckit.companion.* command in order, invoking the real command bodies rather than improvising artifacts, so sizing, design docs, and wave-structured task lists stay faithful to the manual flow. - Automatic Feature Directory Management: Mints a fresh numbered specs/ directory for new work and never writes into a directory that already contains a spec.md, protecting completed specs from being clobbered. - Lifecycle Timing and Checkpoint Recording: Records step, substep, and per-task finish events via write-context.py, sets unattended: true so checkpoint hooks record-and-continue instead of prompting, and ends the spec at status: completed. - Use Case: You describe a new feature in one sentence and let the agent generate the spec, plan, research and data-model docs, contracts, task list, and implementation, then mark the spec complete — all without answering a single review prompt. ## Quick Start Ask the agent to run the full companion pipeline unattended for your feature, for example: run speckit companion auto to build user authentication with passwordless email login.
