zoahdev/dsh-poison-guard

README excerpt

dsh-poison-guard Pre-install supply-chain poison scanner for DeepSeek Harness plugins. It is not a toy regex grep: it runs three layers on every plugin before you dsh plugin add it — 1. AST analysis via NodeSecure JS-X-Ray (the SAST used by NodeSecure CLI): variable tracing, dynamic-import resolution, obfuscator detection, eval / Function / vm sinks, data-exfiltration , serialize-environment , unsafe shell commands, and more. 2. Deobfuscation decoder that unpacks atob() , Buffer.from(..., "base64"/"hex") , String.fromCharCode(...) , and \xNN / \uNNNN escapes, then re-scans the decoded strings for hidden credentials, URLs, and shell commands. 3. Regex heuristics as a fallback for obvious literals, non-code files, and install-time scripts ( prepare / postinstall / install / preinstall ). The honest threat model No static tool can catch all poisoning. Detecting arbitrary malicious behavior in arbitrary code is undecidable (Rice's theorem); a determined attacker can always craft an obfuscation this scanner cannot see through. What this tool does is make the cheap, high-volume attacks — hidden exfiltration URLs, obfuscated require("child process") , eval of base64 blobs, process.env har…

Category