lokalise-deploy-integration

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads the Lokalise CLI binary directly from the official GitHub repository during Netlify build plugin execution and Docker image construction (e.g., https://github.com/lokalise/lokalise-cli-2-go/releases/latest/download/lokalise2_linux_x86_64.tar.gz).
  • [REMOTE_CODE_EXECUTION]: The implementation guides instruct the agent to download, extract, and execute the Lokalise CLI binary within CI/CD and container environments to perform translation synchronization.
  • [COMMAND_EXECUTION]: The skill utilizes shell scripts and Node.js child_process.execSync to manage environment variables, trigger build-time downloads, and execute platform-specific deployment commands (Vercel, Netlify, and GCloud).
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes external content from the Lokalise API.
  • Ingestion points: Fetches translation bundles from api.lokalise.com and GitHub-hosted CLI binaries.
  • Boundary markers: The skill includes integrity checks using jq to validate that downloaded files are valid JSON and that expected language files are present.
  • Capability inventory: Access to the Bash and Write tools, filesystem operations for localized files, and network access for API communication.
  • Sanitization: While it validates JSON structure, it does not sanitize the actual translation strings against potential natural language instructions that could influence downstream agent behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 03:48 AM
Security Audit — agent-trust-hub — lokalise-deploy-integration