The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/setup.sh file fetches the uv installer from https://astral.sh/uv/install.sh. This is a well-known service for Python development tools.
[REMOTE_CODE_EXECUTION]: The scripts/setup.sh script pipes the downloaded installer from astral.sh directly into the shell for execution. This is the standard installation procedure for this trusted tool.
[COMMAND_EXECUTION]: The skill executes local Python scripts and the yt-dlp module to extract document structure and video metadata. These operations are conducted using standard subprocess calls and are essential to the primary task of document conversion and article generation.
[PROMPT_INJECTION]: The skill processes untrusted external content (PDFs, EPUBs, Web URLs, and YouTube transcripts), which presents a surface for indirect prompt injection. * Ingestion points: External data enters the system through scripts/convert_to_markdown.py and scripts/youtube_handler.py. * Boundary markers: The prompts defined in references/article-writer-prompt.md and the tier-specific workflow files use line range delimiters (L{start}-{end}) and section identifiers ([Sxx]) to separate source content from instructions. * Capability inventory: The skill has the capability to spawn subagents via the Task tool and write files to the local disk using the Write tool. * Sanitization: While content is converted to markdown and delimited, no specific sanitization or filtering of adversarial instructions within the source text is performed.

writer-agent