The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill requires the agent to generate CommonJS (.cjs) scripts (e.g., diagram.gen.cjs) based on provided templates to calculate geometric coordinates for advanced chart types like fishbone, flywheel, and treemaps. These scripts are then executed locally using the node runtime. This represents a pattern of runtime script generation and execution for layout logic.
[EXTERNAL_DOWNLOADS]: The skill invokes npx -y @larksuite/whiteboard-cli@^0.2.13 to download and run a rendering tool from the npm registry. This is a dependency on an external package from a well-known service associated with the skill's vendor.
[COMMAND_EXECUTION]: The skill utilizes lark-cli and npx for whiteboard management and rendering. It explicitly directs the agent to execute these commands (e.g., lark-cli --version) to verify availability without requiring user confirmation.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from existing whiteboards via the +export command, which acts as an ingestion surface for untrusted external data.
Ingestion points: Content is retrieved from whiteboards in raw node or code source formats via references/lark-whiteboard-export.md.
Boundary markers: No specific boundary markers or "ignore instructions" delimiters are documented for data retrieved from the whiteboards.
Capability inventory: Includes lark-cli (network/API access, file operations), npx (remote package execution), and node (local script execution).
Sanitization: No explicit sanitization or validation protocols are mentioned for the ingested whiteboard content.

lark-whiteboard