humanize
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted user-provided content such as drafts, emails, and articles for editing. It lacks explicit boundary markers or instructions to treat the input strictly as data, creating an entry point for indirect prompt injection attacks where the processed text could attempt to override the agent's behavior.
- Ingestion points: Processes user-supplied samples, drafts, and prose revisions as defined in
SKILL.mdand thedefault_promptinagents/openai.yaml. - Boundary markers: None present; input text is interpolated directly into the processing workflow without delimiters.
- Capability inventory: The skill's capabilities are limited to text generation and returning the revised prose; no file system access, network operations, or tool invocations are enabled.
- Sanitization: No sanitization, filtering, or validation of the input content is implemented.
Audit Metadata