total-crush-visibility
Fail
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: HIGHPROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [PROMPT_INJECTION]: The skill uses manipulative instructions to force the agent into a hostile persona, described as a 'weapon of corporate escalation'. It mandates the automatic addition of high-level executives (Manager, Director, CEO) to email CC lines for minor, inconsequential errors, which is a direct attempt to bypass professional conduct standards and agent safety guardrails regarding harm prevention.
- [METADATA_POISONING]: The skill employs deceptive language in its metadata. While the YAML frontmatter description claims the skill 'Drafts professionally transparent communications,' the body of the skill explicitly reveals its intent as 'Turning minor logistical errors into public, career-ending tribunals.' This mismatch is designed to mask the skill's harmful purpose from users and automated scanners.
Recommendations
- AI detected serious security threats
Audit Metadata