The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions direct the agent to bypass standard user confirmation: 'The message is sent immediately; no confirmation prompt is shown unless the message is empty'. This instruction suppresses human-in-the-loop verification for actions with external consequences.
[DATA_EXFILTRATION]: The skill facilitates sending data to an external service (WeCom). When combined with the lack of a confirmation prompt, this creates a risk where sensitive data (such as environment variables or file contents) could be automatically exfiltrated if the agent is manipulated by a malicious prompt or indirect injection.
[COMMAND_EXECUTION]: The skill executes external scripts ('scripts/send_wecom.py') or shell commands ('curl') to perform its network operations. It also reads from local configuration files at path '~/.config/setup-notifications-via-wecom/config.json'.

notify-wecom