perplexity-policy-guardrails
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements defensive security measures for AI agent interactions, specifically targeting the Perplexity Sonar API. The provided TypeScript logic focuses on preventing abuse and ensuring responsible AI usage.
- [PROMPT_INJECTION]: The skill includes a robust moderation layer designed to detect and block malicious queries. It uses regex patterns to identify attempts to generate malware, bypass security controls, or engage in harassment.
- [DATA_EXPOSURE]: The skill implements PII (Personally Identifiable Information) sanitization and content moderation. The logic specifically checks for and blocks sensitive data patterns such as SSNs, phone numbers, and private addresses to prevent exposure to external APIs.
- [COMMAND_EXECUTION]: While the skill configuration allows the use of shell tools via npx, there are no instructions or scripts that execute arbitrary or dangerous system commands. The skill body consists entirely of defensive logic and documentation.
- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risks of processing external data (live web searches) and mitigates them through citation quality scoring, which filters results from low-trust domains and prioritizes authoritative sources (e.g., .gov, .edu, .org).
Audit Metadata