The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a Python script scripts/contact_dept_members.py that executes the dws command-line tool using subprocess.run. The script follows security best practices by passing arguments as a list rather than a shell string, effectively mitigating shell injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill processes data returned from the DingTalk API (via the dws tool), such as user names, department names, and friend request remarks. This creates a surface for indirect prompt injection where malicious content in the organization's directory could attempt to influence the agent's behavior.
Ingestion points: Data enters the agent's context through the output of dws contact commands and the contact_dept_members.py script.
Boundary markers: The skill frequently requests the --format json flag, which helps the agent distinguish between data and instructions through structured parsing.
Capability inventory: The skill can read contact details, list department members, manage friend requests, and modify employee information using the dws CLI.
Sanitization: The skill relies on the agent's internal reasoning and structured JSON output to handle external data, though no explicit string sanitization logic is implemented in the provided scripts.

dingtalk-contact