firebase-ai-logic

Pass

Audited by Gen Agent Trust Hub on Mar 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADS
Full Analysis
  • [SAFE]: The skill integrates with official Firebase services and points to verified Google domains for documentation and setup instructions.
  • [EXTERNAL_DOWNLOADS]: Downloads the official Firebase CLI and SDK using the command npm install -g firebase@latest from the public NPM registry.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input such as text prompts and media files (images, audio, video, PDF) through functions like generateText and analyzeImage.
  • Ingestion points: User-provided strings and file objects in references/usage_patterns_web.md.
  • Boundary markers: Not present in the simplified code examples.
  • Capability inventory: Limited to network requests to the Gemini API via the Firebase SDK; no access to local files, system shells, or sensitive environment variables is requested.
  • Sanitization: Not demonstrated in the example snippets, though the use of Firebase App Check is recommended to secure the client-side API usage.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 22, 2026, 04:22 AM
Security Audit — agent-trust-hub — firebase-ai-logic