The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through its processing of external data from Lark Drive.
Ingestion points: Metadata such as file and folder names are retrieved from the Lark API in the list_files function within run.py and presented to the agent.
Boundary markers: There are no delimiters or instructions to ignore embedded commands within the retrieved file data. Furthermore, the SKILL.md instructions explicitly tell the agent to "prohibit interactive menus" and "directly execute the most reasonable operation," which increases the risk of the agent obeying instructions found in file names.
Capability inventory: The skill includes high-impact capabilities such as delete_file (irreversible deletion) and create_folder, as implemented in run.py.
Sanitization: There is no evidence of sanitization, escaping, or validation of the file names or folder tokens before they are processed by the agent or used in subsequent API calls.

drive-file