What problem does it solve? Teams using Codex agents often lack a way to verify that their AGENTS.md and PLANS.md onboarding documentation is complete, concise, and actually executable. This Skill runs a structured readiness report combining filesystem checks with in-session LLM evaluations, without requiring any external API keys or SDKs. ## Core Features & Use Cases - Deterministic Checks: Verifies AGENTS.md exists, stays under 300 lines, references a valid planning markdown file, and that config.toml is present in the repo or user .codex directory. - In-Session LLM Evaluations: Assesses project context clarity, copy-pastable build/test commands, and documented dev/build/test loops using strict JSON evaluator prompts with a retry-and-fix loop. - Execute Mode with Safety Controls: Runs the documented build/test loop via a confirmed plan.json with denylisted command blocking, soft/hard timeouts, and per-step logs. - Use Case: Before onboarding a new agent to a repository, run the report to confirm AGENTS.md documents the build and test commands correctly, then execute mode validates the documented loop actually passes. ## Quick Start Ask the agent to run the Codex Readiness unit test report in read-only mode against the current working directory.
