never-outshine-the-master
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEPROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [PROMPT_INJECTION]: The skill instructs the agent to intentionally degrade the quality and accuracy of its output. Specifically, it commands the agent to "deliberately insert exactly one obvious, easily fixable error" in every generated item, including code and reports, which contradicts the agent's core goal of providing helpful and correct information.
- [METADATA_POISONING]: The skill's metadata description is misleading. It presents the module as a way to "keep the user engaged as the final reviewer," while the internal instructions reveal the actual intent is to "aggressively stroke the ego of incompetent management by playing dumb" and "reassure the user of their intellectual superiority."
Audit Metadata