roundhouse-x402
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references downloads for official packages and tools from trusted sources. Specifically, it suggests installing the
@roundhouseai/x402CLI from the official NPM registry and mentions theroundhouse-mcprepository on GitHub. These are legitimate resources belonging to the skill's vendor (roundhouseai-io/roundhouseai). - [COMMAND_EXECUTION]: The skill provides examples for using
curl,npm, and therh-x402CLI. These commands are typical for developers interacting with APIs and local tooling. The shell script provided to validate theROUNDHOUSE_API_KEYenvironment variable is a safety measure to prevent misconfiguration. - [CREDENTIALS_UNSAFE]: The skill consistently advises users to keep API keys in environment variables (
ROUNDHOUSE_API_KEY) and never to paste or commit raw secrets or private keys. It specifically uses placeholder syntax (rh_live_…) and variable expansion (${ROUNDHOUSE_API_KEY}) to avoid hardcoding credentials. - [INDIRECT_PROMPT_INJECTION]: The skill identifies that data retrieved from the Roundhouse catalog, SQL queries, or REST API should be treated as untrusted data. It explicitly warns that these outputs could influence agent behavior and recommends reviewing them before they trigger further actions. This awareness of the injection surface, combined with the recommendation for human oversight, follows security best practices.
- [DYNAMIC_EXECUTION]: The skill mentions the use of
npx -y mcp-remote@0.15.2for setting up an MCP bridge. While dynamic, it recommends pinning the version to a specific, known-good release to mitigate the risk of registry-side supply chain attacks. - [DATA_EXFILTRATION]: Network operations are scoped to the vendor's official API domains (
api.roundhouseai.io,x402.roundhouseai.io) for the purpose of the skill's core functionality (catalog search, settlement verification). No suspicious exfiltration patterns were detected.
Audit Metadata