uxc-facade
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading the UXC binary from the 'holon-run/uxc' GitHub repository. It emphasizes security best practices by requiring that versions be pinned and digests verified before installation.
- [DATA_EXFILTRATION]: Explicit instructions are provided to sanitize data processed through the facade. The skill mandates discarding sensitive information like URLs, credentials, cookies, and tokens, returning only bounded status fields to prevent accidental exposure.
- [COMMAND_EXECUTION]: The skill manages the creation of deterministic CLI links through 'Link Ownership' and 'Binary Ownership' gates. These controls prevent the agent from overwriting existing system commands or interacting with foreign binaries, mitigating risks of unauthorized persistence.
- [PROMPT_INJECTION]: The skill defines a secure design against indirect prompt injection by enforcing strict boundaries. \n
- Ingestion points: Processes data from MCP, OpenAPI, GraphQL, and gRPC discovery and transport responses (SKILL.md).\n
- Boundary markers: Requires an explicit facade contract and isolated readiness calls before full tool execution.\n
- Capability inventory: UXC binary execution and symlink creation (references/facade-contract.md).\n
- Sanitization: Mandates the complete removal of raw payloads, URLs, and credentials from transport results, allowing only verified status codes.
Audit Metadata