firebase-ai-logic-basics
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is authored by Firebase (Google LLC) and provides implementation patterns for official, well-known SDKs and tools. All external links and packages are hosted on trusted domains (firebase.google.com) or registries (npm), representing standard and safe development practices for the Firebase ecosystem.
- [SAFE]: The documentation includes robust security guidance regarding secret management. It provides specific patterns for Web, Android, and iOS to avoid hardcoding or committing App Check debug tokens to version control, recommending the use of environment variables and gitignored local configuration files instead.
- [INDIRECT_PROMPT_INJECTION]: The skill describes patterns for processing untrusted external data, such as user-provided images, audio, and documents, which represents an indirect prompt injection surface for applications built using these instructions.
- Ingestion points: Untrusted data is ingested via multimodal inputs (images, audio, video, PDFs) in the
generateContentandgenerateContentStreammethods across all platforms. - Boundary markers: The documentation includes instructions for configuring safety thresholds (
SafetySetting) to block harmful content generated from these inputs. - Capability inventory: Applications built with these patterns can perform network operations (Firebase API calls) and execute local functions via the 'Function Calling' feature.
- Sanitization: The skill relies on Firebase's server-side safety filters and App Check for request attestation, which are industry-standard mitigations.
Audit Metadata