hermes-cn-desktop
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documentation includes instructions to install the official Rust toolchain using the standard installer from rustup.rs.
- [COMMAND_EXECUTION]: Provides instructions for local development and build tasks, including standard package management commands (pnpm, cargo) and a suggestion to modify file permissions (chmod +x) for the local runtime binary during troubleshooting.
- [DYNAMIC_EXECUTION]: The Rust code snippets illustrate how the application manages its background runtime using subprocess spawning (std::process::Command) and describes a mechanism for downloading and verifying runtime updates.
- [INDIRECT_PROMPT_INJECTION]: The skill describes an application designed to process and display data from external LLM providers, which creates a potential surface for indirect prompt injection if the processed content contains malicious instructions.
- Ingestion points: The application processes chat responses and streaming data through its frontend API clients and UI components.
- Boundary markers: None explicitly shown in the provided code snippets for handling untrusted LLM output.
- Capability inventory: The backend includes capabilities for network proxying, subprocess management, and local file system interaction.
- Sanitization: Zod schemas are used to validate configuration data, though no specific sanitization is shown for streaming message content.
Audit Metadata