serpapi-reference-architecture

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests data from external search engine results via SerpApi, creating a potential surface for indirect prompt injection attacks.
  • Ingestion points: Data is ingested from the SerpApi client within the SearchService.search and SearchService.trackKeyword methods in SKILL.md.
  • Boundary markers: The reference implementation lacks explicit boundary markers or instructions to the model to ignore any instructions found within the ingested search content.
  • Capability inventory: The service demonstrates the capability to write to a database (db.saveRanking) using the data retrieved from the external source.
  • Sanitization: No sanitization or validation logic is shown in the processing of external organic results before they are used by the system.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 08:40 AM
Security Audit — agent-trust-hub — serpapi-reference-architecture