pi-agent-harness
Pass
Audited by Gen Agent Trust Hub on Jul 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a technical manual for the Pi Agent Harness development environment. It includes instructions for building, testing, and extending the agent framework.
- [COMMAND_EXECUTION]: Describes the use of local development commands such as
npm run check,npm install, and the agent's built-in tools likebashandedit. These are standard requirements for a coding assistant framework. - [EXTERNAL_DOWNLOADS]: Documents the framework's ability to install extensions and packages from remote sources like npm and GitHub via
pi install. This is a core functional feature of the software being documented. - [CREDENTIALS_UNSAFE]: References the use of environment variables for managing LLM provider API keys (e.g.,
ANTHROPIC_API_KEY), which is a standard and recommended practice for such tools. No hardcoded credentials or unauthorized exfiltration patterns were found. - [DATA_EXFILTRATION]: Mentions a
/sharecommand for uploading session data to GitHub Gists. This is a documented, user-triggered feature for session sharing and does not represent hidden exfiltration.
Audit Metadata