The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill automatically installs its required Python and Node.js dependencies, including reportlab, pypdf, matplotlib, and playwright, from official package registries if they are not detected on the system. It also fetches font assets from Google's well-known font service (fonts.googleapis.com) during the document cover rendering process.
[COMMAND_EXECUTION]: Shell commands are invoked within scripts/make.sh and through helper functions in several Python scripts (scripts/fill_inspect.py, scripts/merge.py, scripts/render_body.py, etc.) using subprocess.check_call. These executions are used for dependency verification, package installation, and managing the PDF generation pipeline, and they utilize hardcoded package names and system paths.
[PROMPT_INJECTION]: The skill processes user-supplied documents (Markdown, PDF, Text, JSON) to reformat or fill them, which introduces an indirect prompt injection vulnerability surface.
Ingestion points: External files are read and parsed by scripts/reformat_parse.py and scripts/fill_inspect.py.
Boundary markers: The skill does not employ specific delimiters or clear instructions to the agent to ignore any embedded commands within the content being processed.
Capability inventory: The skill scripts have the capability to write to the local filesystem, perform network operations for asset retrieval, and execute shell commands for maintenance and installation.
Sanitization: While the content is converted into a structured JSON schema (content.json) before the rendering phase, the skill lacks specific security-focused sanitization to detect or neutralize malicious instructions that may be present in the original input text.

minimax-pdf