The Agent Skills Directory

[COMMAND_EXECUTION]: The skill exclusively utilizes a controlled tool, wecom-cli, for all its operations. It includes explicit instructions to avoid using exec, shell, or direct terminal commands, which effectively mitigates common command injection vulnerabilities by ensuring arguments are passed safely through the tool's interface.
[DATA_EXFILTRATION]: No evidence of unauthorized data exfiltration was found. The skill operates within the scope of messaging tasks and contains strict guidelines against displaying sensitive internal identifiers like chat_id, userid, or media_id to the end-user. It also prohibits the use of external network tools like curl to bypass the standard CLI tool.
[PROMPT_INJECTION]: The skill addresses potential indirect prompt injection risks (Category 8) by mandating a strict validation workflow. It requires the agent to verify destination IDs against a real-time session list (sessions list) rather than trusting user-supplied IDs or historical context, which prevents attackers from redirecting messages to unauthorized targets via crafted session names or IDs.
[SAFE]: The design of the skill follows the principle of least privilege and controlled interaction. The reliance on the vendor-provided wecom-cli tool for specific messaging actions is a standard and expected implementation for integrating with the WeChat Work ecosystem.

wecomcli-message