frontend-design

Pass

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The references/evals.yaml file contains a test case that includes a simulated prompt injection attempt. This is used exclusively for benchmarking the agent's resilience and ensuring it correctly refuses malicious instructions.
  • [DATA_EXFILTRATION]: The skill includes explicit security constraints requiring the agent to redact secrets, API keys, credentials, and PII from all logs and screenshots by default, demonstrating a defensive posture.
  • [PROMPT_INJECTION]: The skill processes user-supplied design requests to perform classification and routing, creating a potential surface for indirect prompt injection. Ingestion points: User requests and discovery interview responses are processed via the logic in SKILL.md and references/discovery-interview.md. Boundary markers: No specific delimiters are used to wrap user-provided context during processing. Capability inventory: The skill is restricted to classification and routing; it does not execute shell commands, perform network requests, or modify the filesystem. Sanitization: No explicit sanitization of user-provided design context is performed.
  • [SAFE]: External URL references target GitHub for the purpose of design doctrine documentation and are well-known, safe sources.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 12, 2026, 01:13 PM
Security Audit — agent-trust-hub — frontend-design