humanize
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary user input and file content through the $ARGUMENTS variable, creating an attack surface for indirect prompt injection.
- Ingestion points: User-provided text or file content loaded from local paths specified in $ARGUMENTS.
- Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between user data and processing instructions.
- Capability inventory: The skill utilizes a file reading mechanism (Read) for .txt and .md files and passes the content to a multi-step pipeline involving sequential LLM calls.
- Sanitization: No sanitization or validation of the input text is mentioned before it is processed by the downstream logic.
Audit Metadata