a2a-setup
Pass
Audited by Gen Agent Trust Hub on Mar 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads the OpenClaw A2A Gateway plugin from the author's GitHub repository (win4r/openclaw-a2a-gateway).
- [EXTERNAL_DOWNLOADS]: Fetches the official installation script from Tailscale, a well-known networking service, to establish connectivity between servers.
- [REMOTE_CODE_EXECUTION]: Executes the Tailscale installation script by piping the remote content directly to the shell (curl | sh), which is the standard installation method for this service.
- [REMOTE_CODE_EXECUTION]: Uses the Python
json.toolmodule to format JSON data retrieved via curl from a local endpoint for verification purposes. - [COMMAND_EXECUTION]: Performs several configuration tasks using the
openclawCLI,npm installfor dependency management, andopensslfor local generation of security tokens. - [COMMAND_EXECUTION]: Provides an attack surface for indirect prompt injection (Category 8) by creating a communication bridge between agents.
- Ingestion points: The gateway receives incoming JSON-RPC and REST messages from peer agents.
- Boundary markers: None explicitly implemented in the provided TOOLS.md template.
- Capability inventory: The agent uses the
exectool to run thea2a-send.mjsscript for outbound communication. - Sanitization: The skill relies on the underlying @a2a-js/sdk and OpenClaw platform for data handling.
Audit Metadata