The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local bash script (scripts/dt_helper.sh) to manage DingTalk API credentials and tokens. It also executes generated scripts from the /tmp directory to interact with DingTalk endpoints.
[DYNAMIC_EXECUTION]: The skill instructions direct the agent to generate multi-line shell scripts at runtime, save them to temporary files (e.g., /tmp/<task>.sh), and execute them to perform calendar operations as described in the workflow section of SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data which is interpolated into API request bodies, creating a vulnerability surface.
Ingestion points: User-provided event summaries, descriptions, and participant identifiers are collected and processed within SKILL.md for calendar event creation and modification.
Boundary markers: Absent; the provided execution templates do not include specific delimiters or 'ignore' instructions to prevent the agent from obeying instructions potentially embedded in user-supplied calendar data.
Capability inventory: The skill possesses file-write capabilities (to /tmp), shell execution (bash), and network access via curl to DingTalk's API domains.
Sanitization: No sanitization or escaping mechanisms are specified in the instructions for user-provided strings before they are interpolated into shell commands or JSON payloads.

dingtalk-calendar