The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to retrieve and process user-generated content from the Teambition API, which serves as a potential vector for indirect prompt injection attacks.
Ingestion points: Scripts such as query_tasks.py, query_task_detail.py, query_task_activity.py, and render_rtf.py ingest external data including task titles, notes, and comments.
Boundary markers: While SKILL.md provides detailed formatting and display rules, it lacks explicit boundary markers or instructions to the agent to disregard instructions potentially embedded within the retrieved task data.
Capability inventory: The skill possesses capabilities for network operations (API requests to Teambition) and file uploads to the vendor's storage infrastructure via upload_file.py.
Sanitization: The scripts process API responses into structured JSON or simplified text for the agent; however, the agent remains exposed to the raw text content of retrieved comments and task notes.
[DATA_EXFILTRATION]: The skill manages sensitive data, including a TEAMBITION_USER_TOKEN. All network operations are directed towards official vendor endpoints (open.teambition.com) or temporary presigned URLs provided by the official API for file uploads, which is a standard and expected behavior for this integration.
[COMMAND_EXECUTION]: The skill leverages uv and python3 to execute its internal scripts as defined in the package.json and README.md. No patterns suggesting the execution of arbitrary or unsanitized shell commands from untrusted inputs were identified.

dingtalk-teambition