The Agent Skills Directory

[SAFE]: The skill documents a legitimate design workflow using the op CLI and MCP tools. It focuses on PenNode schema, semantic roles, and layout rules for vector design.
[COMMAND_EXECUTION]: The instructions include examples of shell commands for interacting with the op CLI. This includes using python3 as a utility to parse JSON IDs from command output, which is a standard development practice for CLI-based workflows.
[PROMPT_INJECTION]: The skill processes design data from various sources (JSON strings, DSL, SVG, and Figma files), which creates an attack surface for indirect prompt injection. \n
Ingestion points: op open, op import:figma, op import:svg, op design, and op insert (SKILL.md). \n
Boundary markers: None specified for user-provided data. \n
Capability inventory: The op CLI provides capabilities for file system access (op save, op theme:save), design manipulation, and submission of generated code chunks (op codegen:submit). \n
Sanitization: No specific sanitization or validation of the content of imported files is described.

openpencil-design