The Agent Skills Directory

[SAFE]: The skill provides a legitimate and well-documented workflow for technical illustration, using modern software development practices.
[EXTERNAL_DOWNLOADS]: The scripts/setup.sh script fetches font files from jsDelivr and GitHub repositories. These operations are performed using hardcoded SHA256 integrity hashes to ensure the downloaded assets have not been tampered with.
[COMMAND_EXECUTION]: The toolchain utilizes standard CLI tools (npm, uv, node, python) and automates a headless browser via the Playwright library. These capabilities are used strictly for their intended purposes: dependency management, environment setup, and graphic rendering.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data such as graph structures in JSON and chart data in Python scripts. An evidence chain for this category is established:
Ingestion points: scripts/layout.mjs (reads graph configurations) and scripts/figure_mpl.py (reads chart data).
Boundary markers: The workflow enforces a FigureSpec and automated linting.
Capability inventory: File system writes and subprocess execution are used for rendering.
Sanitization: The built-in linting system (scripts/lint.mjs) automatically scans generated SVG files for potentially dangerous or non-portable elements such as <script> tags, foreignObject, or unauthorized external references in the figure source, mitigating the risk of malicious content entering the rendering pipeline.

technical-figure