canva-core-workflow-b
Pass
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill exclusively communicates with official Canva domains (api.canva.com) using standard authentication patterns.
- [COMMAND_EXECUTION]: The skill uses
readFileSyncto read local files for the purpose of uploading them to Canva as design assets. This is a functional requirement for the skill's primary purpose. - [EXTERNAL_DOWNLOADS]: The skill supports uploading assets via URL, which utilizes Canva's backend to fetch content from external sources.
- [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface as it processes template metadata and field names from the Canva API and interpolates user-provided text into design fields.
- Ingestion points: Canva API responses from
/brand-templatesand/datasetendpoints, as well as user-supplied text for autofilling templates. - Boundary markers: Not present.
- Capability inventory: Local file reading, network request capability to Canva APIs, and restricted shell access (npm).
- Sanitization: None observed; the skill relies on the underlying agent's handling of data strings.
Audit Metadata