The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the required dws command-line utility through Python scripts (scripts/todo_batch_create.py, scripts/todo_daily_summary.py, and scripts/todo_overdue_check.py) using subprocess.run. This is the primary intended mechanism for the skill to interact with the DingTalk service. The scripts include robust safety measures, such as input validation and a mandatory confirmation process for batch operations using a SHA-256 plan digest to ensure user-approved changes are not modified before execution.
[INDIRECT_PROMPT_INJECTION]: The skill handles untrusted external data, including task titles, descriptions, and comments, which are processed via the dws todo CLI. This creates a potential surface for indirect prompt injection if the ingested content contains instructions for the agent. However, the risk is mitigated by the skill's strict enforcement of structured JSON output, clear boundary markers for data processing, and explicit instructions requiring human confirmation for all write actions.

dingtalk-todo