The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a local script (scripts/render.ts) to execute the agent-browser tool. This is used to open a temporary HTML wrapper and capture screenshots of the generated SVG for visual verification, which is an intended and safe part of the illustration workflow.
[DYNAMIC_EXECUTION]: The instructions encourage the agent to write and run temporary scratch scripts to generate SVG path data and mathematical wavering effects. This code generation is confined to the creation of static image files and does not involve executing untrusted remote code.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and interpret the prose of documentation articles to create visual metaphors. This creates a surface where instructions embedded in the article text could influence the agent's artistic choices. However, the impact is limited to the visual content of the generated SVG.
Ingestion points: Content of the documentation articles being illustrated (SKILL.md, Part 1).
Boundary markers: No specific delimiters are used to separate article content from agent instructions.
Capability inventory: File system writes (SVG files), execution of the local rendering script, and use of the agent-browser and view_image tools.
Sanitization: No specific sanitization of article text is performed, as the input is used for creative conceptualization rather than command construction.

editorial-illustration