bro
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes previous assistant messages which may contain data from untrusted sources.\n
- Ingestion points: The skill body in
SKILL.mdidentifies the "previous assistant response" as the data source to be restated.\n - Boundary markers: The instructions provide behavioral constraints such as "Do not introduce new claims" and "Preserve the meaning," though it lacks structural delimiters for the input data.\n
- Capability inventory: The skill definition states that no tools or files are needed, and no
allowed-toolsare specified in the frontmatter.\n - Sanitization: There is no evidence of input filtering or sanitization before the data is processed.
Audit Metadata