The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to its document and comment processing features.
Ingestion points: Untrusted data enters the agent context through the read and list_blocks actions for document content, as well as list_comments and get_comment actions for document feedback in SKILL.md.
Boundary markers: Absent. The skill instructions do not define delimiters or provide 'ignore embedded instructions' warnings for the agent when handling content retrieved from documents or comments.
Capability inventory: The skill provides extensive write and delete permissions, including write, create_and_write, append, update_block, delete_block, and create_comment actions.
Sanitization: Absent. There are no indications of content filtering, escaping, or validation for the data fetched via the Feishu API before it is presented to the agent's context.

feishu-doc