perplexity

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface. Ingestion points: The scripts perplexity_ask.py, perplexity_reason.py, perplexity_research.py, and perplexity_search.py ingest data via command-line arguments. Boundary markers: Absent. User input is passed to the model without delimiters. Capability inventory: Scripts call router.invoke to perform network-based AI queries. Sanitization: Absent. No filtering is performed on user input.
  • [EXTERNAL_DOWNLOADS]: Dependency on vendor resources. The skill uses the ai_agent_marketplace package and deepnlp.org endpoints, which are legitimate resources for the author aiagenta2z.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 08:45 AM
Security Audit — agent-trust-hub — perplexity