codex-fix

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). At runtime, the skill invokes the Codex plugin subagent via the Task tool (subagent_type: "codex:codex-rescue"), and the subagent’s returned findings (findings_raw) are fed into the agent context for the subsequent hand-off to refactor-verify; this is outsider-authored free text produced by an external plugin/model rather than the operating user.