The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the agent to install the playwright framework and download chromium browser binaries. It also fetches several frontend libraries (p5.js, Chart.js, Leaflet, Plotly, Mermaid, etc.) from public CDNs like unpkg.com and jsdelivr.net.
[COMMAND_EXECUTION]: The generator workflow involves running numerous local Python utilities and shell scripts for tasks such as extracting simulation specs, scaffolding files, validating quality, and capturing screenshots. Some of these scripts are intended to be installed into the user's local binary path (~/.local/bin).
[INDIRECT_PROMPT_INJECTION]: The skill includes a utility (extract-sim-specs.py) that parses textbook chapter source files (index.md) to extract parameters for simulation generation. This creates a surface where malicious content in the processed documents could influence the resulting code.
Ingestion point: extract-sim-specs.py reads content from index.md files.
Boundary markers: Uses markdown headers (#### Diagram:) as structural delimiters, but lacks adversarial injection defenses.
Capability inventory: The skill has extensive filesystem write access and can generate/modify executable scripts.
Sanitization: No sanitization of user-provided specifications is performed during the extraction or scaffolding phases.
[DYNAMIC_EXECUTION]: The skill is primarily a code generator that produces HTML and JavaScript. Additionally, it implements a 'Docker Python Lab' pattern where Python code is sent to a local HTTP service for execution, representing a surface for dynamic code evaluation.

microsim-generator