perplexity-reference-architecture

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill ingests untrusted user data which is then interpolated into prompts sent to the Perplexity API.
  • Ingestion points: The 'topic' parameter in ResearchOrchestrator.research and the 'claim' parameter in factCheck in both SKILL.md and implementation.md.
  • Boundary markers: The implementation uses standard double quotes as delimiters which do not provide robust protection against escaping.
  • Capability inventory: The skill performs network operations via the openai client to api.perplexity.ai.
  • Sanitization: No sanitization or escaping of user-provided strings is implemented.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 09:39 PM
Security Audit — agent-trust-hub — perplexity-reference-architecture