research-decision-room
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted research material to generate an interactive HTML decision room, creating a potential surface for indirect prompt injection where malicious data could influence the agent's output or execute in the user's browser.
- Ingestion points: The
research_materialinput inSKILL.mdaccepts arbitrary strings from external sources such as support tickets, survey excerpts, and interview notes. - Boundary markers: The instructions do not define clear delimiters or "ignore" markers to separate the ingested untrusted content from the skill's own instructions.
- Capability inventory: The skill utilizes the
file_writecapability to emit the synthesized data into anindex.htmlartifact. - Sanitization: There are no explicit instructions for the agent to sanitize, escape, or filter HTML/JavaScript within the
research_materialbefore interpolating it into the final HTML output.
Audit Metadata