perplexity-deploy-integration
Pass
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: SAFE
Full Analysis
- [CREDENTIALS_SAFE]: The skill demonstrates best practices for secret management, utilizing platform-specific tools like
vercel envandgcloud secretsto handle thePERPLEXITY_API_KEYrather than hardcoding sensitive data. - [INDIRECT_PROMPT_INJECTION]: The provided code templates create API endpoints that ingest untrusted user input via the
queryparameter and pass it to the Perplexity Sonar API. This represents a standard integration surface for LLM-powered applications. - Ingestion points:
req.json()inapi/search.tsandreq.bodyinserver.ts(SKILL.md). - Boundary markers: Absent in the simplified code templates.
- Capability inventory: The skill uses
Bashtools for production deployments to Vercel and Google Cloud Run. - Sanitization: No specific input sanitization is implemented in the provided examples.
Audit Metadata