genkit-production-expert
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands to manage the application lifecycle, including project initialization via
scripts/init-genkit.shand deployment usingfirebaseandgcloudCLIs. - [EXTERNAL_DOWNLOADS]: Fetches official Google Cloud and Firebase development libraries (e.g.,
@genkit-ai/googleai,firebase-genkit) from standard package registries. - [INDIRECT_PROMPT_INJECTION]: The skill generates AI flows that process external data, creating an attack surface for indirect prompt injection. This is mitigated through documented best practices:
- Ingestion points: Runtime input schemas defined in
references/examples.md(e.g.,QuestionInput,RagInput). - Boundary markers: System instructions in generated prompts guide the model to stick to provided context.
- Capability inventory: The skill utilizes file writing and bash command execution for project management.
- Sanitization: Enforces strict runtime validation using Zod and Pydantic, and provides specific error handling for
SAFETY_BLOCKresponses. - [SAFE]: Credentials are managed via Google Cloud Secret Manager rather than being hardcoded, adhering to standard security best practices.
Audit Metadata