kim-decision
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its architecture of ingesting untrusted user data into complex internal reasoning frameworks (Meta_Kim discipline) which then drive tool usage. Malicious instructions in user-provided data could potentially manipulate the agent's decision-making logic or tool interactions.
- Ingestion points: Data enters through user chat and the request_user_input tool as specified in SKILL.md.
- Boundary markers: The instructions do not specify any delimiters (such as specific tags or markers) to separate user-provided data from the agent's internal reasoning instructions.
- Capability inventory: The skill instructions in references/gates.md authorize the use of web search, file reading, and API query tools to verify evidence.
- Sanitization: There are no requirements or instructions for the agent to sanitize, filter, or escape user-provided content before it is processed by the reasoning framework.
Audit Metadata