The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external, potentially untrusted documents (PDF, Office, and images), which constitutes an indirect prompt injection surface.
Ingestion points: The skill uses parser.parse(file_path) and lit parse to process files, as seen in SKILL.md and scripts/batch_parse_dir.py.
Boundary markers: No explicit instructions are provided to the agent to treat the output of the parsing tool with specific delimiters or as untrusted content.
Capability inventory: The skill utilizes Read, Write, Edit, and Bash tools to manage files and execute parsing commands.
Sanitization: Data is processed locally using the liteparse library; no remote LLM calls are used for the extraction process.
[EXTERNAL_DOWNLOADS]: The skill references the installation of liteparse and @llamaindex/liteparse from official package registries (PyPI and npm). These are well-known services and the packages are associated with a recognized organization (LlamaIndex).
[COMMAND_EXECUTION]: The skill utilizes local command-line tools including the lit CLI, soffice (LibreOffice), and magick (ImageMagick) to perform document conversion and parsing tasks. This execution is limited to local operations essential to the skill's primary function.

liteparse