style-mimic
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided text samples, creating an attack surface where malicious instructions could be embedded to influence agent behavior.
- Ingestion points: Untrusted data enters the agent context through the writing sample variables in
SKILL.md(<sample_1>, <sample_2>, and <sample_3>). - Boundary markers: The template uses XML-like delimiters to isolate samples, but lacks instructions for the agent to disregard potential commands found within the samples.
- Capability inventory: No active capabilities such as file writes, network access, or shell execution are present across the skill's instructions.
- Sanitization: No sanitization or filtering logic is implemented for the writing samples.
Audit Metadata