The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Canva:start-editing-transaction followed by Canva:cancel-editing-transaction to retrieve design metadata such as element positions and sizes. This process is used strictly for data inspection and ensures no changes are persisted to the user's design.\n- [PROMPT_INJECTION]: The skill processes untrusted design content, presenting a surface for indirect prompt injection. Untrusted data enters the context via Canva:get-design-content (ingestion point). No explicit boundary markers or sanitization steps are defined in the instructions. However, the skill's capabilities are limited to read-only tool calls within the Canva environment (capability inventory), and the agent is directed to provide structured feedback, which mitigates the risk of executing instructions found within design elements.\n- [SAFE]: Data retrieval for thumbnails and design content is performed through authorized Canva API tools. The workflow includes a design resolution step to ensure the agent operates on correctly identified design resources.

canva-design-feedback