The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the dws command-line utility to perform searches for personnel, enterprise knowledge, and behavioral records within a corporate environment.
Evidence: Multiple command templates such as dws aisearch person --keyword "<keyword>" --dimension <dimension> and dws aisearch enterprise --queries "<queries>" --types <types> are defined for agent use.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests and processes untrusted enterprise data such as chat messages, emails, and documents.
Ingestion points: Data returned from dws aisearch enterprise (IM, mail, documents) and dws aisearch behavior (interaction records).
Boundary markers: The skill refers to dws-shared for global safety rules and preflight checks, but does not explicitly define per-command delimiters for search results within these files.
Capability inventory: The agent can write to other skills (chat, DING) using retrieved data and execute local CLI commands.
Sanitization: No explicit sanitization or filtering logic is defined within the instruction files for handling retrieved content before it is processed by the agent.

dingtalk-aisearch