skills/agentscope-ai/qwenpaw/mailbox/Gen Agent Trust Hub

mailbox

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted email content from external providers, creating a surface for malicious instructions to influence agent behavior.\n
  • Ingestion points: External data enters the context via list_messages, get_message, search_messages, and get_thread as described in SKILL.md.\n
  • Boundary markers: The instructions mandate a manual confirmation step from the user before the agent performs sensitive actions such as delete_message or delete_thread.\n
  • Capability inventory: The skill allows for high-impact write operations including send_message, reply_message, and delete_message in SKILL.md.\n
  • Sanitization: The skill relies on human oversight rather than automated filtering to prevent the execution of malicious instructions found within emails.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 09:52 AM
Security Audit — agent-trust-hub — mailbox