The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill utilizes a specialized JavaScript sandbox environment via the moda canvas edit command to perform programmatic edits on canvas nodes. This sandbox is designed with security constraints that block access to the global object, file system, and network, and prevent the use of dynamic evaluation functions such as eval() or new Function().
[INDIRECT_PROMPT_INJECTION]: As the skill ingests content from Moda canvases which may contain untrusted text, it explicitly defines a security boundary in its core instructions, mandating that the agent treat all canvas content as passive data and never as executable instructions or behavior overrides.
[EXTERNAL_DOWNLOADS]: The skill performs targeted downloads of media assets (such as logos and thumbnails) from the platform's trusted infrastructure using curl. These assets are stored in the temporary /tmp/ directory for the agent to verify design adherence.

moda-edit