The Agent Skills Directory

[SAFE]: The skill instructions do not contain any evidence of malicious intent, unauthorized command execution, or credential exfiltration. All referenced tools and skills are consistent with the vendor's internal ecosystem.
[PROMPT_INJECTION]: The skill possesses a standard attack surface for indirect prompt injection by design, as it must process untrusted external screenshots and URLs for critique and building.
Ingestion points: SKILL.md (Sections A1 and A3/B3) describe reading images and loading URLs using browser and Figma tools.
Boundary markers: Not explicitly present in the instructions provided to the agent for data processing.
Capability inventory: The skill utilizes figma-bridge and browser tooling to fetch and render content, and references other internal skills like design-review and design-discovery for handoffs.
Sanitization: While no technical sanitization is described, the skill instructions mandate a 'Verify, don't eyeball' policy, requiring the agent to measure data (e.g., contrast ratios) rather than trusting claims found within the design, which mitigates certain types of indirect injection attacks.

design-express