goal-ready-spec
Goal-Ready Spec
Overview
Create file-backed specs that an execution agent can follow without replacing explicit requirements with merely equivalent behavior. The goal is to collaborate with the user until the spec is clear, executable, and faithful to their intent, not to finish a spec in one pass. The output must make architectural ownership, state location, scope boundaries, verification commands, and completion evidence auditable before any goal is marked complete. Plan files should be compact-resilient and context-efficient: keep spec.md as a small router and controlling contract, keep hot-path files focused on current state, and route detailed phase, context, audit, and evidence material into smaller files that are opened only when needed. The final response should hand the user a concise prompt that asks Codex to set its own goal from the generated spec file, using the goal-design advice from OpenAI's "Using Goals in Codex" cookbook: define the outcome, evidence surface, constraints, boundaries, iteration policy, and terminal condition.
Do not use this skill for ordinary planning, brainstorming, outlining, refining, reviewing, or implementation-plan requests unless the user explicitly invokes $goal-ready-spec or specifically asks to make the spec goal-ready.