The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates and executes plotting code to render quantitative figures from user-supplied data.
Evidence: The 'Plotter' role brief in references/agent-prompts.md explicitly instructs the agent to "Write and run the plotting code."
[DYNAMIC_EXECUTION]: The skill uses the execute_code tool to interact with the Penpot API for advanced operations like SVG export.
Evidence: references/penpot-redraw.md notes that SVG export can be performed via execute_code using shape.export({type: "svg", scale: 1}).
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data which is subsequently interpolated into instructions for planning, styling, and image generation.
Ingestion points: User-provided source material and intent are recorded in brief.md as described in SKILL.md.
Boundary markers: The skill utilizes structured role briefs in references/agent-prompts.md to scope agent tasks, though it does not define specific delimiters for separating user content from instructions.
Capability inventory: The skill possesses code execution capabilities through the Plotter role and the Penpot MCP integration.
Sanitization: Instructions emphasize following the user's supplied references as authoritative but do not detail specific sanitization or escaping of external content.

scientific-illustrations