goal

What problem does it solve? It prevents agents from starting work on vague or ambiguous requests by forcing every goal through a structured intake that produces a written, verifiable contract defining what "done" means before any execution begins. ## Core Features & Use Cases - Structured intake and clarification: Reads project memory, asks one clarifying question at a time, and classifies the goal into a five-tuple (taxonomy, execution mode, completion mode, template, confidence) with lifecycle-aware defaults. - Verifiable contract generation: Writes a human-readable contract.md plus a machine-readable contract.json sidecar containing weighted acceptance criteria, verification methods, budgets, and a definition of done, then registers the goal as ACTIVE only after human approval. - Optional research mode: With the --research flag, runs market-validation or technical research via WebSearch/WebFetch, writes research.md, and gates new-capability ideas behind an explicit go/no-go/modify decision. - Use Case: A founder says "add a billing page." The skill clarifies scope and success criteria, writes a contract with measurable acceptance criteria, gets approval, registers the goal, and hands off to /plan and /loop for execution. ## Quick Start Invoke /goal with a description of what you want to build, optionally adding --research to validate a new idea against the market before the contract is written.