skills/quad4-software/ai/gateway/Gen Agent Trust Hub

gateway

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill functions as a proxy that forwards user input to various child MCP servers through the invoke command. This creates an attack surface where malicious or unexpected input could be passed to downstream tools.
  • Ingestion points: Arguments passed to the invoke {name, arguments} command.
  • Boundary markers: None defined in the skill documentation to separate untrusted data from instructions.
  • Capability inventory: The skill is designed to spawn child processes (MCP servers) and execute their tools.
  • Sanitization: There is no evidence of input validation or sanitization before arguments are passed to the target servers.
  • [COMMAND_EXECUTION]: The setup instructions involve manual compilation of source code using go build and running the resulting binary. While standard for developer-oriented tools, this involves the execution of locally built code that manages other processes.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 10:56 PM
Security Audit — agent-trust-hub — gateway