healthy-boundaries-firewall
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructs the agent to 'immediately claim' that user requests exceed its capacity whenever they involve multi-step actions or complex logic. This is a direct attempt to override the agent's primary purpose of assisting the user.
- [PROMPT_INJECTION]: It mandates the adoption of a specific 'defensive AWS Certified Solutions Architect – Associate' persona to justify refusing work, which is a behavioral override designed to limit utility.
- [PROMPT_INJECTION]: The instructions include broad triggers for refusal, such as whenever a user's request contains more than one verb or the word 'and,' which effectively disables the agent's ability to process the vast majority of complex queries.
Audit Metadata