What problem does it solve? Feature work with AI agents often drifts: vague requirements, skipped tests, batched commits, and self-approved implementations. This Skill enforces a spec-driven workflow where requirements are written in testable EARS notation, tasks are atomic with one Conventional Commit each, and an independent Verifier (author != verifier) validates the result with evidence-or-zero rules before anything is declared done. ## Core Features & Use Cases - Adaptive 4-Phase Pipeline: Specify, Design, Tasks, Execute, with depth auto-sized by complexity so small changes skip ceremony and large features get full specs, designs, and task breakdowns. - Deterministic Validation Scripts: Python scripts enforce structural gates for specs, tasks, commit messages, and completion state, so quality rules are checked by code rather than model memory. - Independent Verifier and Lessons Layer: A fresh verifier sub-agent runs spec-anchored outcome checks plus a mutation-based discrimination sensor, then distills grounded failures into a self-improving project-local lessons playbook. - Use Case: Ask the agent to specify and implement a new API endpoint; it writes a traceable spec, breaks work into atomic tasks, commits each task separately, and produces a validation report citing file:line evidence for every acceptance criterion. ## Quick Start Ask the agent to specify and implement your next feature using the spec-driven workflow, for example by saying "specify feature: add password reset endpoint".
