The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill generates custom interactive visualizations via an app_block component that uses HTML and vanilla JavaScript derived from data inputs. While this is necessary for functionality, it creates a potential vulnerability surface.
Ingestion points: Data is ingested from reviewed analytical results as specified in SKILL.md and native-inline-visualizations.md.
Boundary markers: The skill uses specific platform control sequences (genui and ) as UI delimiters and instructs the agent to keep data 'embedded and bounded'.
Capability inventory: The skill possesses the capability to generate and execute Python code (Matplotlib) and render interactive HTML/JS/SVG components.
Sanitization: The skill instructions include security-hardening measures by strictly prohibiting inline event handlers, iframes, external scripts, and network requests during the rendering process.
[Dynamic Execution]: The skill dynamically assembles JSON specifications for platform-native charts and generates plotting code for Matplotlib images at runtime. This behavior is standard for visualization tools and follows structured templates intended to limit arbitrary execution.

visualize-data