rivet-agentos
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill contains extensive examples (e.g.,
examples/processes/exec.ts,examples/bindings/guest.py) demonstrating how agents can execute shell commands and manage process lifecycles. These operations occur within isolated virtual machines and are the primary intended functionality of the platform. - [EXTERNAL_DOWNLOADS]: The examples facilitate the integration of official software packages (e.g.,
@agentos-software/pi,@agentos-software/claude-code) and recognized third-party tools like the Browserbase CLI. These are standard components for building agent capabilities within the agentOS ecosystem. - [CREDENTIALS_UNSAFE]: The skill promotes secure development by providing patterns for injecting API keys and secrets into agent sessions using environment variables at runtime (as seen in
examples/llm-credentials/client.ts), which avoids the risks associated with hardcoded credentials.
Audit Metadata