The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on shell scripts (bin/mineru-cli, scripts/*.sh) and a pre-packaged virtual environment to execute the mineru binary. These operations are core to the tool's functionality for converting PDF documents but involve the execution of bundled code on the user's system.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted PDF files and converts them into Markdown for the AI agent to read without any sanitization or boundary markers.
Ingestion points: PDF documents provided by the user and processed via the convert or batch commands.
Boundary markers: Absent; the skill does not wrap the converted Markdown output in delimiters or provide instructions to the agent to ignore embedded commands.
Capability inventory: The skill can write files to the local file system and execute shell commands through its scripts and CLI utility.
Sanitization: No validation, escaping, or filtering is performed on the text content extracted from the PDF files before it is presented to the agent.
[SAFE]: The skill uses and references the MinerU open-source engine from the well-known opendatalab organization for its PDF processing tasks.

mineru-converter