The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a local CLI binary (canva) to interact with the Canva MCP server, perform tool discovery, and execute design actions.
[EXTERNAL_DOWNLOADS]: Fetches design thumbnails from Canva's official servers and supports importing assets from public HTTPS URLs. These operations target a well-known service and are consistent with the skill's primary purpose.
[DATA_EXFILTRATION]: Provides functionality to upload local files to Canva's platform using the upload-file command. While this involves reading local data and sending it to a remote service, it is a core feature of the design tool and requires user approval as defined in the manifest.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple sources, creating a potential attack surface. 1. Ingestion points: User-provided design briefs in create-design, external assets imported via URL in upload-asset-from-url, and image thumbnails downloaded via curl. 2. Boundary markers: The instructions specify using defined tool schemas and providing a user_intent for uploads, though explicit text delimiters for the briefs are not mentioned. 3. Capability inventory: The skill can execute CLI commands, write to the workspace/ directory, and perform network requests to Canva's API. 4. Sanitization: No explicit sanitization or filtering logic is described for the processed text or image content.

canva