sc-research
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill documentation explicitly lists the use of
mcp__rube__RUBE_REMOTE_BASH_TOOLand a localBashtool for processing research results and managing large response data. While listed for functional purposes, these provide a high-privilege execution environment.\n- [INDIRECT_PROMPT_INJECTION]: The skill architecture relies on scraping arbitrary web content and passing it to analysis models, which is a known vector for indirect prompt injection attacks.\n - Ingestion points: Web search queries and URL content extraction (scraping) identified in Phase 3.\n
- Boundary markers: No specific delimiters or safety instructions are defined to isolate scraped data from the agent's core instructions.\n
- Capability inventory: The skill has access to
mcp__rube__RUBE_REMOTE_BASH_TOOL,Bash, andWrite(file system access) as specified in the MCP Integration and Tool Coordination sections.\n - Sanitization: There is no evidence of sanitization, validation, or filtering of external web content before it is processed by the analysis and consensus tools.
Audit Metadata