The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from SVG and HTML files which could contain malicious instructions designed to influence the agent.
Ingestion points: The scripts/check_diagram.py script reads diagram content using a headless browser via Playwright.
Boundary markers: Text extracted from diagrams is printed directly to the console in error reports and wording correction findings without delimiters or protective instructions.
Capability inventory: The skill can write files (PNG screenshots) and provides detailed textual feedback that becomes part of the agent's context.
Sanitization: No sanitization or escaping is applied to text content extracted from the diagrams before it is printed.
[COMMAND_EXECUTION]: The scripts/test_check_diagram.py script uses subprocess.run to execute the diagram checker during local automated testing. This is standard behavior for a validation suite.
[DYNAMIC_EXECUTION]: The scripts/check_diagram.py script uses Playwright's evaluate method to execute a local JavaScript file (diagram_geometry.js) within a browser context for geometric analysis. This execution uses a static, resolved path to the local source file.
[EXTERNAL_DOWNLOADS]: The skill depends on the playwright library, which is a well-known service for browser automation. The dependency is declared in a PEP 723 script block within the checker script.

make-diagram