anti-ai-slop-writing
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEMETADATA_POISONINGINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill documentation and banned words list contain references to future dates and reports (e.g., 'Carnegie Mellon 2025', 'March 2026 report', 'GPT-5 era 2025 onward') which are factually deceptive and misrepresent the skill's source material.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-supplied data across a wide range of writing tasks (emails, articles, bios, etc.) but lacks explicit boundary markers or sanitization, creating a surface for indirect prompt injection.
- [PROMPT_INJECTION]: The directive to 'Apply all rules silently' and 'Never mention them' is a concealment pattern that hides the skill's influence from the user's view, reducing oversight of the agent's behavior.
Audit Metadata