Codex Result Handling

This is internal plumbing that formats Codex static analysis output for human consumption. It enforces a critical safety rule: after showing review findings, Claude must stop and explicitly ask which issues you want fixed before touching any code. No auto-applying fixes, even obvious ones. It also preserves the structure Codex returns (severity ordering, file paths, line numbers, evidence boundaries) and handles edge cases like failed runs, malformed output, and auth errors without improvising workarounds. You won't invoke this directly, but it's why Claude presents security and correctness findings as a menu instead of silently rewriting your codebase.