The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/init-firebase.sh script automates the installation of the official firebase-tools CLI via npm install -g firebase-tools. Additionally, references/implementation.md provides CI/CD configurations that utilize official GitHub Actions from the actions/ and FirebaseExtended/ organizations.- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for project initialization, dependency installation (npm install), and deployment (firebase deploy). This is consistent with its primary purpose as a deployment and setup tool.- [DATA_EXFILTRATION]: Examples in references/examples.md and references/implementation.md use curl for connectivity and health checks targeting local emulator ports or project-specific hosting URLs. These operations are used for deployment verification and do not exfiltrate sensitive data to external domains.- [PROMPT_INJECTION]: The skill provides templates for RAG (Retrieval-Augmented Generation) and content moderation that interpolate data from Firestore into LLM prompts. This creates an indirect prompt injection surface if the Firestore data is user-controlled.\n
Ingestion points: Untrusted data enters the context through Firestore document triggers (e.g., documents/{docId}, posts/{postId}) as seen in references/examples.md.\n
Boundary markers: The example prompts do not implement delimiters or specific instructions to ignore embedded commands within the ingested text.\n
Capability inventory: The generated code has the capability to perform Firestore reads/writes via firebase-admin and execute AI inference via @google-cloud/vertexai (as documented in references/examples.md and ARD.md).\n
Sanitization: The provided code examples do not include explicit sanitization or filtering logic for the data before it is sent to the model.

firebase-vertex-ai