linkfox-zhihuiya-patent-forward-citation

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill and its associated scripts interact exclusively with vendor-controlled domains belonging to linkfox.com and linkfox-ai.com. This includes API gateways, authentication services, and feedback endpoints.
  • [SAFE]: Sensitive data such as API keys are managed through standard environment variables (e.g., LINKFOX_AGENT_API_KEY), avoiding hardcoded secrets within the codebase.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external patent citation data from a third-party API (Zhihuiya), which creates an indirect prompt injection surface. However, this is the primary purpose of the skill and no dangerous capability-chaining was detected.
  • Ingestion points: Patent citation and literature data fetched via scripts/zhihuiya_cited_references.py.
  • Boundary markers: Data is organized into structured JSON and presented in Markdown tables as defined in SKILL.md.
  • Capability inventory: The skill writes API responses to the local project directory and provides textual summaries to the agent context.
  • Sanitization: Responses are parsed as JSON to ensure structure, though specific NLP filtering for the content itself is not explicitly implemented.
  • [PERSISTENCE_MECHANISMS]: The onboarding script provides instructions for users to persist their API keys in shell profiles (~/.zshrc, ~/.bashrc). This is a standard and transparent setup procedure for development tools and does not constitute a hidden persistence threat.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 07:03 AM
Security Audit — agent-trust-hub — linkfox-zhihuiya-patent-forward-citation