adobe-migration-deep-dive
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external workflow data and third-party platform documentation, creating a surface for potential instruction injection. The skill mitigates this by requiring verification against first-party documentation and establishing strict human-in-the-loop controls.
- Ingestion points: Analysis of source workflow inventories, data formats, and Adobe API documentation files.
- Boundary markers: Explicitly requires named product, security, and data owners to approve specific migration waves and live cutovers.
- Capability inventory: The skill uses Read, Glob, Grep, Write, and Edit tools to inspect and generate repository artifacts; network access is explicitly restricted from skill invocation.
- Sanitization: Requires the use of synthetic fixtures with secret and content canaries to validate behavior in non-production environments.
- [EXTERNAL_DOWNLOADS]: The skill references multiple resources from the official Adobe developer portal for authentication and API guidance.
- Evidence: Links are provided to developer.adobe.com for OAuth, Firefly, PDF Services, and Photoshop API documentation to ensure the agent uses current, verified service contracts.
Audit Metadata