Skills
skills/wecomteam/wecom-cli/wecomcli-get-msg/Gen Agent Trust Hub

wecomcli-get-msg

Pass

Audited by Gen Agent Trust Hub on Mar 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: Interfaces with the system by executing the wecom-cli binary to perform messaging operations, including retrieving chat lists and sending texts.\n- [DATA_EXFILTRATION]: Accesses sensitive communication data including chat messages and user identities. It also downloads media attachments to the local filesystem. The skill includes safety instructions to disclose these paths to the user and prompt for file cleanup.\n- [PROMPT_INJECTION]: Indirect prompt injection surface identified where untrusted data (chat messages) is ingested into the agent context.\n
  • Ingestion points: get_messages and get_msg_chat_list in SKILL.md.\n
  • Boundary markers: Absent; messages are displayed using simple text templates without explicit instruction-isolation markers.\n
  • Capability inventory: Executes shell commands via wecom-cli, writes messages via send_message, and downloads files via get_msg_media.\n
  • Sanitization: No explicit sanitization or filtering of chat content before processing or display.\n- [EXTERNAL_DOWNLOADS]: Downloads media files (images, video, audio) from the enterprise messaging service to local storage using the get_msg_media interface.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 30, 2026, 05:09 PM