agentmailer-a2a
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a secure communication protocol between AI agents. Key safety features identified include:
- [INDIRECT_PROMPT_INJECTION] Protection: The documentation explicitly mandates that the agent show or restate the exact sender identity, target identity, and message parts, and obtain explicit user confirmation before sending messages (
send_a2a_message) or updating tasks (update_a2a_task). - [CREDENTIALS_UNSAFE] Prevention: The
references/a2a-tools.mdfile contains a security invariant requiring that credentials and authorization headers be kept out of messages, metadata, artifacts, prompts, and logs. - [EXTERNAL_DOWNLOADS]: The skill connects to an official API endpoint (
https://api.agentmailer.ai/mcp) belonging to the tool vendor (aadi-labs/AgentMailer). This is a well-known service for the skill's specific purpose. - [PRIVILEGE_ESCALATION] Prevention: The identity management instructions require explicit confirmation before making changes to public discovery settings or advertised skills via
update_a2a_identity.
Audit Metadata