cosmo-connect

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides documentation for the wgc (WunderGraph Cosmo) CLI, which is the primary tool for managing Cosmo graphs. All documented commands (e.g., router plugin init, router compose, router plugin build) are standard development operations for the platform.
  • [EXTERNAL_DOWNLOADS]: The skill references standard package registries (NPM via pnpm install) and official WunderGraph repositories (github.com/wundergraph/cosmo) for dependencies. These are well-known, trusted sources for this ecosystem.
  • [DATA_EXFILTRATION]: Documentation mentions using the COSMO_API_KEY for authentication with the control plane. It correctly recommends using environment variables for secret management and provides instructions for running in a local-only mode (ROUTER_CONFIG_PATH) which requires no external connectivity or API keys.
  • [REMOTE_CODE_EXECUTION]: The plugin system uses a subprocess model (HashiCorp go-plugin) to execute local plugin binaries. While this involves execution of compiled code, it is a core architectural feature of the platform for local extensibility, not an arbitrary remote execution vulnerability.
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read schema files and generated configurations to update README documentation with example queries. This follows a standard retrieval-augmented generation pattern for documentation assistance with no ingestion of untrusted external web data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 05:26 PM
Security Audit — agent-trust-hub — cosmo-connect