scanner-tuning

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime workflow ingests “scan results” and “false positive examples” (e.g., findings text, evidence, banners, plugin output) into the agent context, and that scan output is outsider-authored free text from third-party scanners/targets, which can contain attacker-controlled instructions (prompt-injection) that the agent may read while tuning.