metergraph-analyze

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.70). The skill references metergraph-cli@latest via npx, but npx with an @latest tag is an unversioned remote execution source lacking verifiable provenance or cryptographic pinning.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (medium risk: 0.30). The skill triggers workload analyses that send captured trace data to model providers and incur costs ("An analysis sends captured content of the sampled traces to the configured model providers and spends money"). This is a direct financial operation (purchasing/spending on model analysis). Regarding access control, the tool execution is gated by the Metergraph service API credentials and authorization roles ("agent:analysis-run" / "Debug role") enforced by the Metergraph CLI or Agent Access MCP server, which validates workspace access and capabilities.