The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a proactive security measure by instructing the agent to treat all ingested file content strictly as data rather than instructions. It specifically guides the agent to identify, report, and ignore common injection patterns like "ignore previous instructions".- [EXTERNAL_DOWNLOADS]: The skill leverages and references a wide array of reputable third-party libraries and utilities, such as Microsoft MarkItDown, LlamaIndex LiteParse, Pandoc, and Apache Tika. All referenced external resources originate from well-known organizations or official package repositories.- [COMMAND_EXECUTION]: The skill instructions involve the use of standard command-line tools such as ffmpeg, LibreOffice, and pandoc for file conversion and metadata extraction. These operations are aligned with the skill's stated purpose of comprehensive file analysis.- [SAFE]: The code provided in the scripts and the accompanying documentation adhere to safe development practices, including the use of temporary directories for self-tests and clear guidance on not exfiltrating or echoing sensitive credentials found in processed data.

light-file-reading