perplexity

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses subprocess.run to orchestrate between its Python and Node.js components, specifically calling scripts/chrome-token.mjs to handle browser-based authentication.
  • [EXTERNAL_DOWNLOADS]: Declares dependencies on external packages @rezkam/browser-tools (from the NPM registry) and perplexity-webui-scraper (from PyPI) to provide browser automation and API scraping capabilities.
  • [PROMPT_INJECTION]: As the skill ingests and summarizes web search results and LLM outputs from Perplexity, it maintains an indirect prompt injection surface where content from external websites could attempt to influence the agent's instructions.
  • [DATA_EXFILTRATION]: The skill extracts the __Secure-next-auth.session-token cookie from the user's local Chrome profile. This is the primary and documented authentication method used to authorize requests sent to perplexity.ai.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 09:30 AM
Security Audit — agent-trust-hub — perplexity