developing-genkit-go
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to install the Genkit CLI using a shell script from the official Genkit domain.\n
- Evidence:
curl -sL cli.genkit.dev | bashinSKILL.md.\n- [INDIRECT_PROMPT_INJECTION]: The skill provides guidance on building agents and generative UI that process untrusted data, which is an inherent risk factor for indirect prompt injection. The documentation proactively addresses this by including security warnings and mitigation strategies.\n - Ingestion points: Built agents ingest data from user messages and external tool outputs (e.g., in
SKILL.md,references/agents.md).\n - Boundary markers: The documentation recommends using
aix.ValidateResumeAgainstHistoryand notes that protocol validation is not a substitute for a security boundary (e.g., inreferences/a2ui.md,references/agents-custom.md).\n - Capability inventory: The Genkit library provides tools for file system access (
Filesystemmiddleware), network operations (model providers), and session state management.\n - Sanitization: The guidance advises on the importance of client-side prop sanitization and the use of Content Security Policies (CSP) to restrict remote sources (e.g., in
references/a2ui.md).
Audit Metadata