humanizer
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external evidence and scraped data using the sandbase_run tool, which introduces a vulnerability surface where instructions embedded in untrusted web content could be processed by the agent.
- Ingestion points: External data sources, search results, and web scraping content retrieved through the sandbase_run API tool (SKILL.md).
- Boundary markers: The skill lacks instructions for using delimiters or explicit 'ignore embedded instructions' prompts when handling retrieved external data.
- Capability inventory: The agent can invoke multiple vendor API tools (sandbase_discover, sandbase_inspect, sandbase_run) and generate implementation artifacts on the local system.
- Sanitization: There are no requirements for sanitizing or filtering external content before it is interpolated into the agent's context for humanization tasks.
Audit Metadata