The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates the ingestion of external documents (PDFs, images, etc.) which are parsed into Markdown. This creates a surface for potential indirect prompt injection from malicious document content. This is an inherent characteristic of document processing tools.
Ingestion points: client.parse in SKILL.md accepts local paths or remote document_url inputs.
Boundary markers: Extracted text is not explicitly delimited with instructions to ignore embedded commands.
Capability inventory: Includes network API requests and local file writing for saving outputs.
Sanitization: Extracted Markdown is returned as-is without content filtering.
[EXTERNAL_DOWNLOADS]: The tool can download documents from remote URLs and retrieve processed data from presigned S3 URLs. These operations are essential to the skill's functionality and target the vendor's infrastructure or user-specified locations.
[COMMAND_EXECUTION]: The documentation includes standard setup commands for installing Python dependencies (landingai-ade, python-dotenv, pymupdf) and initializing environment variables.

document-extraction