executor-capability-gate

What problem does it solve? Calling external AI models like Codex or Gemini without pre-flight checks wastes time on failed invocations caused by missing CLIs, expired logins, network outages, or rate limits. This Skill runs five checks before every external model call so failures are caught upfront. ## Core Features & Use Cases - Five Pre-Flight Checks: Verifies CLI installation, login credentials (e.g., ~/.codex/auth.json), network connectivity via curl, local rate-limit timestamps, and task suitability for external execution. - Structured Gate Output: Produces a pass/fail report per check and decides whether to dispatch to Codex/Gemini or fall back to Claude execution. - Result Caching: Caches gate results for 5 minutes per session to avoid redundant checks, with cross-platform stat handling for Linux and macOS. - Use Case: When running a code review with the [C] parameter to delegate to Codex, the gate detects that Codex was called only 30 seconds ago and the task touches authentication code, so it automatically falls back to Claude instead of failing mid-review. ## Quick Start Before dispatching any task to Codex or Gemini, run the executor capability gate checks and tell me whether the external call is allowed or should fall back to Claude.