create-agent
Warn
Audited by Snyk on Jun 20, 2026
Risk Level: MEDIUM
Full Analysis
MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).
- Potentially malicious external URL detected (high risk: 0.90). The skill includes runtime dependencies that will be contacted/pulled during execution — notably the container image pkg.harness.io/vrvdt5ius7uwygso8s0bia/harness-agents/claude-code-plugin:main (pulled and executed at runtime) and external MCP endpoints used by the agent such as https://api.githubcopilot.com/mcp/ and the ngrok example https://harmfully-unregulative-theressa.ngrok-free.dev/mcp (and placeholder https:///mcp) which the agent calls at runtime to perform actions — these endpoints/images therefore execute remote code or directly enable remote control of agent actions.
Issues (1)
W012
MEDIUMUnverifiable external dependency detected (runtime URL that controls agent).
Audit Metadata