What problem does it solve? Turning vague proposals or tasks into clear, testable requirements is error-prone, and AI agents often drift, hallucinate scope, or lose state between sessions. This Skill distills a proposal into a concise, testable SPEC.md with stable R-NNN requirement identifiers and persists workflow state in STATE.json so any agent can resume the workstream later. ## Core Features & Use Cases - Spec Generation: Distills user proposals, codebase context, and AGENTS.md constraints into observable, implementation-neutral requirements tagged with unique R-NNN identifiers. - Revision Workflow: Rewrites SPEC.md from scratch after a failed verification, converting every VERIFY.md hard block into an addressed spec item. - State Persistence: Creates and updates .spae/[workstream]/STATE.json, advancing the phase from spec to plan so downstream agents can resume with zero prior context. - Use Case: You have a feature idea like "add rate limiting to the API." Invoke the skill to produce a scoped SPEC.md with requirements, testing strategy, out-of-scope boundaries, and assumptions, ready for the plan phase to decompose into atomic tasks. ## Quick Start Ask the agent to run the spae-spec skill with your proposal, for example: create a spec for adding rate limiting to the API endpoints.
