minimax-mcporter-tools
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill's MiniMax.understand_image tool is designed to accept absolute and relative local file paths as image sources. While this is necessary for its stated purpose of image analysis, it provides the agent with a capability to access the local file system and transmit data to a third-party service.
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it incorporates untrusted data from external sources (web search results and image contents) into the agent's context without sanitization. 1. Ingestion points: Organic search results from MiniMax.web_search and extracted text or descriptions from MiniMax.understand_image. 2. Boundary markers: Absent; the skill does not instruct the agent to use delimiters or to ignore embedded instructions within tool outputs. 3. Capability inventory: The skill can perform network lookups and read local files. 4. Sanitization: Absent; there is no evidence of filtering or escaping logic for the retrieved external data.
Audit Metadata