The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted document content from external wiki instances, which creates an indirect prompt injection surface if documents contain malicious instructions intended to manipulate the AI agent.
Ingestion points: Untrusted text is retrieved from the Outline Wiki API endpoints (/documents.info and /documents.search) and output via scripts/outline.py in the read and search command handlers.
Boundary markers: Content is framed by basic text lines (--- Content ---) but lacks robust structural boundaries or instruction overrides to signal to the agent that the content should be treated strictly as passive data.
Capability inventory: The tool has capabilities to create and update remote wiki documents, as well as save exported files locally via the export command's --output argument.
Sanitization: No sanitization, validation, or prompt-escaping mechanisms are performed on the retrieved document body before it is processed by the agent.

outline