Track live agent benchmark validation for Codex CLI.\n\nCommand target:

ctx leaderboard --hallucination --live --agent codex

Acceptance:

• Run in an external environment where Codex CLI auth/session is available.
• Capture skipped/error output if the CLI is unavailable or blocked.
• Do not use this result in launch copy until it is reproducible.
• Keep offline deterministic benchmark as the primary launch claim.