ShellWard
A compliance and security scanning tool for AI applications, built around Chinese regulations, scanning a project in one command for foreign model endpoints, hardcoded keys and exposed personal data, with runtime interception of prompt injection.
ShellWard is a compliance gateway for AI applications aimed at Chinese regulatory requirements: the Cybersecurity Law, the Personal Information Protection Law, the Multi-Level Protection Scheme, cross-border data transfer rules, and AI content labelling. A single command scans a project, read-only and without uploading data, and prints a scorecard with findings down to the exact file and line, such as a foreign large-model endpoint, a hardcoded token, or a phone number in a data file.
A companion agent skill for tools such as Claude Code, Cursor or Codex runs a deeper compliance audit: it starts from the scan's deterministic baseline, traces how personal data flows through the code, and produces a report where every finding must cite an exact file and line that a verification script checks, and every finding is challenged once by an independent review agent. Facts the repository cannot establish are surfaced as a question for a person rather than assumed.
It can run as a browser-based scanner over a public repository URL, as a local web interface for private code that never leaves the machine, or in CI with a JSON or HTML report and a failing exit code on critical findings, and it suggests domestic model alternatives such as Qwen, DeepSeek, Kimi or Zhipu when it detects a foreign model dependency.