genkit

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill recommends installing the Genkit CLI using curl -sL cli.genkit.dev | bash. This is a documented installation pattern for an official Google/Firebase tool and targets a well-known service domain.\n- [COMMAND_EXECUTION]: The skill documents various CLI commands (e.g., genkit start, genkit flow:run) for interacting with the Genkit environment and executing AI flows during development and testing.\n- [PROMPT_INJECTION]: As a framework for building RAG (Retrieval-Augmented Generation) and agentic systems, the skill demonstrates patterns that ingest untrusted user input and external data, creating a surface for indirect prompt injection.\n
  • Ingestion points: User-provided text in summarizeFlow, agentFlow, and ragFlow flows; external documents retrieved via devLocalVectorstoreRetriever.\n
  • Boundary markers: Code examples use system prompts like 'Answer questions using only the provided context' to guide the LLM, though strict delimiter-based enforcement is left to the developer.\n
  • Capability inventory: Flows are designed to execute arbitrary logic, call external tools/APIs, and interact with databases.\n
  • Sanitization: The framework strongly encourages the use of Zod schemas for input and output validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 12:43 PM