The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data (PDFs, images, and scanned documents) through OCR, creating a surface for indirect prompt injection.
Ingestion points: External files provided via file_urls, file_ids, or content_base64 (SKILL.md).
Boundary markers: The instructions lack explicit delimiters or warnings to the agent to disregard instructions embedded within the extracted OCR text.
Capability inventory: The skill can extract full text and structured entities, which are then returned to the agent's context.
Sanitization: No evidence of text sanitization or filtering to prevent the agent from executing instructions found in the processed documents.
[EXTERNAL_DOWNLOADS]: The skill documentation includes instructions to install related dependencies from the author's repository.
Evidence: npx skills add AgentPMT/agent-skills --skill what-is-agentpmt (SKILL.md).
Source: These downloads target the official AgentPMT repository, consistent with the skill's authorship.

document-ocr-agent