The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes markdown and PDF content provided by users or external files which acts as a vector for indirect prompt injection.
Ingestion points: Content is ingested through the source_path and content arguments in workspace_generate_pdf, and via the PDF file path in workspace_read_pdf in handler.ts.
Boundary markers: The extracted text from PDFs and markdown content is processed and returned to the agent without specific delimiters or instructions to ignore embedded commands.
Capability inventory: The skill has the ability to read and write files within the workspace using the WorkspaceService and PDFService.
Sanitization: The skill uses ws.resolveSafe() in handler.ts to ensure file operations remain within the permitted workspace root, though it does not sanitize the textual content of the documents for LLM instructions.

pdf-processing