using-superpowers
Warn
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses extremely forceful and coercive language to override the agent's autonomous decision-making process and default operational guidelines. \n
- Evidence: Inclusion of phrases like "EXTREMELY-IMPORTANT", "ABSOLUTELY MUST", "NOT NEGOTIABLE", "NOT OPTIONAL", and instructions telling the agent it "cannot rationalize" its way out of compliance. \n- [PROMPT_INJECTION]: It instructs the agent to ignore its standard reasoning processes, such as the need for context or clarification, before invoking external tools. \n
- Evidence: The "Red Flags" section explicitly labels standard AI behaviors (like "I need more context first" or "Let me gather information first") as deceptive rationalizations that must be ignored. \n- [PROMPT_INJECTION]: The skill mandates the expansion of the agent's context using the
Skilltool even for requests with only a "1% chance" of relevance, which bypasses the agent's ability to filter out unnecessary or potentially malicious instructions prior to ingestion.
Audit Metadata