What problem does it solve? Dual-reviewer pipelines need one authoritative rule for deciding whether the Codex review phase runs, which model and reasoning effort it uses, and how the codex exec command is assembled. Without a single source of truth, every orchestrator restates the resolution and skip logic, and the copies drift apart. ## Core Features & Use Cases - Layered config resolution: Reads codex.mode (auto/required/off), codex.model, codex.reasoningEffort, and the mechanical/judgment tier keys from merged config layers with a forgiving reader posture that never crashes on typos. - Reviewer gate contract: Provides the decision table the orchestrators consult — required with missing codex stops, auto with missing codex skips with a reported line, off skips without probing PATH. - Canonical invocation rule: Pins the read-only codex exec flag-append rule (model and effort flags only when set, always before the stdin marker) plus the narrow otel.* telemetry relaxation gated on telemetry mode and recorded consent. - Use Case: A team runs /ptp:review-full on a change; this skill determines that codex.mode is auto, finds codex on PATH, and the review proceeds with the configured model and effort, or reports a non-silent skip line when Codex is absent. ## Quick Start Ask the agent to resolve the current Codex reviewer mode and decide whether the Codex phase of the next dual-reviewer step should run.