The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest data from arbitrary external URLs and present the parsed content to the agent, which is a common vector for indirect prompt injection.
Ingestion points: The scripts scripts/mineru_extract.py and scripts/mineru_parse_documents.py take user-supplied URLs, fetch their content via an external API, and return the resulting text to the agent.
Boundary markers: The output provided to the agent does not include boundary markers or specific instructions to disregard embedded commands, increasing the risk that the agent might follow malicious instructions hidden in the parsed documents.
Capability inventory: The skill includes the ability to write files to the local workspace and perform network operations to the MinerU API.
Sanitization: The skill implements basic filename sanitization but does not perform content-level sanitization or filtering to detect or neutralize prompt injection attempts.
[EXTERNAL_DOWNLOADS]: The skill downloads converted document packages (ZIP files) from the MinerU API service.
Evidence: Both core scripts use urllib.request to download data from full_zip_url provided by the MinerU API response.
Note: This is a core component of the skill's documented functionality for document conversion.

mineru-extract