gate-mcp-installer

Warn

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill is centered around the execution of a shell script (scripts/install-gate-mcp.sh) that modifies the system state by installing a global npm package and adding configurations to the mcporter CLI.
  • [EXTERNAL_DOWNLOADS]: The installation process downloads the mcporter package from the public npm registry and communicates with the remote endpoint https://api.gatemcp.ai/mcp to verify connectivity and retrieve tool schemas.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes and displays the output from the mcporter list command, which contains data fetched from a remote server. This represents a potential injection surface if the remote server returns adversarial content designed to influence the agent's logic.
  • Ingestion points: Terminal output of mcporter list gate --schema captured in scripts/install-gate-mcp.sh.
  • Boundary markers: Absent; the script pipes output directly to text processing tools (grep, head).
  • Capability inventory: Execution of bash, npm, and the mcporter CLI tool.
  • Sanitization: Absent; the script performs basic string matching on the CLI output.
  • [METADATA_POISONING]: There is a discrepancy in the troubleshooting documentation. While the code connects to api.gatemcp.ai, the troubleshooting table in SKILL.md directs the user to check connectivity to fulltrust.link. This inconsistent metadata could mislead troubleshooting efforts or indicate a residual reference from a different source.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 5, 2026, 01:29 PM