The Agent Skills Directory

[DYNAMIC_EXECUTION]: The open-pencil eval command utilizes a -c flag to execute arbitrary JavaScript strings within the design engine's context. This capability is used for scripting and automation but allows for the execution of dynamically provided code logic.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external .fig files. Since these files can contain arbitrary text data and the MCP server provides tools to read this data into an AI agent's context, there is a risk of indirect prompt injection if the processed design content contains malicious instructions.
Ingestion points: Reads local .fig files via CLI and MCP server tools (tree, find, node, query, eval, and MCP read tools).
Boundary markers: None specified in the documentation for delimiting design content from instructions.
Capability inventory: The skill has broad capabilities including reading/writing files (-w flag, 90 MCP tools) and executing JavaScript code via the eval CLI command.
Sanitization: Not explicitly documented for text content read from Figma nodes.
[COMMAND_EXECUTION]: The documentation includes instructions for running the MCP server using npx openpencil-mcp, which involves downloading and executing packages from the registry at runtime.
[EXTERNAL_DOWNLOADS]: The installation instructions reference external sources for software acquisition, including a Homebrew tap (open-pencil/tap), NPM/Bun packages (@open-pencil/cli, @open-pencil/mcp), and direct GitHub repository cloning.

open-pencil-design-editor