firebase-cli

Pass

Audited by Gen Agent Trust Hub on Sep 27, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The scripts/install.sh script and SKILL.md instructions include the command curl -sL https://firebase.tools | bash. This fetches the official Firebase CLI installation script from a well-known service and executes it directly in the shell to facilitate standalone installation without Node.js.
  • [EXTERNAL_DOWNLOADS]: The skill uses npm install -g firebase-tools within scripts/install.sh and SKILL.md to install the required command-line utilities from the official npm registry. This is the primary and recommended installation method for the package.
  • [COMMAND_EXECUTION]: Helper scripts located in scripts/deploy.sh and scripts/emulators.sh utilize the eval command to run firebase CLI operations. These commands are dynamically constructed using arguments such as project identifiers, deployment targets, and feature flags passed during invocation.
  • [INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for processing untrusted data as it translates natural language deployment and configuration requests into CLI commands.
  • Ingestion points: User-provided deployment messages, project aliases, and feature configurations processed by the agent to generate shell commands.
  • Boundary markers: None present; the skill relies on the agent's internal safety filters.
  • Capability inventory: The skill can install global packages, write configuration files (firebase.json, .firebaserc), execute shell commands, and perform network deployments via the Firebase CLI.
  • Sanitization: Input parameters are interpolated into shell command strings within the bash scripts without dedicated escaping or validation logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 27, 2026, 08:47 PM