The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided data for chart and infographic generation, creating a surface for indirect prompt injection where instructions embedded in the data could influence the agent's behavior.
Ingestion points: Data provided by the user for charts, graphs, and infographics (SKILL.md).
Boundary markers: None identified in the provided instructions; there are no explicit 'ignore embedded instructions' warnings for the data processing phase.
Capability inventory: The skill uses tools (generate_chart, create_visual_design) that execute arbitrary Python code via a code interpreter environment. This includes file-writing capabilities (plt.savefig, canvas.save, image.save) and potential network or system access depending on the execution environment's restrictions (SKILL.md, canvas-design.md).
Sanitization: No evidence of sanitization, escaping, or validation of user-provided data before it is incorporated into the code generation process.
[DYNAMIC_EXECUTION]: The skill relies on generating and executing Python code at runtime using libraries like matplotlib, reportlab, Pillow, and svgwrite. While these are used for visual design, the ability to generate and execute arbitrary code represents a significant capability (SKILL.md, library-reference.md).
[COMMAND_EXECUTION]: The diagram-design.md file suggests using os.remove('temp.svg') for cleanup. While benign in this context, it demonstrates the agent's ability to execute shell-level file system commands.

visual-design