The Agent Skills Directory

[SAFE]: The skill provides integration for DingTalk Todo using the vendor-provided 'dws' CLI tool. No malicious patterns such as prompt injection, unauthorized data exfiltration, or persistence mechanisms were detected.
[COMMAND_EXECUTION]: Several Python scripts (todo_batch_create.py, todo_daily_summary.py, todo_overdue_check.py) execute shell commands using the subprocess module. Analysis shows these are safe as they use list-based arguments instead of shell strings, effectively preventing shell injection. Additionally, the scripts implement timeouts and input validation for parameters like priority and recurrence rules.
[DATA_EXFILTRATION]: While the skill includes an 'add-attachment' command that can upload local files, this is a legitimate functionality of the product. No suspicious network operations or attempts to access sensitive system files (e.g., SSH keys, credentials) were found.
[INDIRECT_PROMPT_INJECTION]: The skill includes a batch processing feature (todo_batch_create.py) that reads from a JSON file.
Ingestion points: scripts/todo_batch_create.py reads a user-specified JSON file.
Boundary markers: Not explicitly defined in the file format, but the data is processed as structured JSON.
Capability inventory: Subprocess calls to the 'dws' CLI tool.
Sanitization: The script validates data types and value ranges for priority and recurrence fields. It uses list-based subprocess calls to ensure inputs are treated as arguments rather than executable shell code. Given these controls and the intended use case, the risk is negligible.

dingtalk-todo