The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill's workflow explicitly instructs the agent to create temporary shell scripts at /tmp/<task>.sh and execute them using bash. This runtime script generation is used to perform API operations, handle token caching, and process variables while avoiding shell-related truncation issues.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data, such as task titles, descriptions, and participant identifiers, which are then interpolated into the generated shell scripts and JSON payloads.
Ingestion points: User input for ToDo tasks (subjects, descriptions) processed via the instructions in SKILL.md.
Boundary markers: None present to distinguish user data from command logic within the generated scripts.
Capability inventory: The skill utilizes shell execution (bash), file writes (/tmp/, ~/.dingtalk-skills/config), and network operations (curl).
Sanitization: No explicit sanitization or escaping mechanisms are defined to handle malicious characters in user input before interpolation into shell commands.
[COMMAND_EXECUTION]: The skill uses various command-line tools (grep, sed, curl, bash) to manage configuration files and communicate with external APIs. These commands handle sensitive information, including AppKeys and Access Tokens, which are stored in the user's home directory.

dingtalk-todo