SECURITY SIGNALS

Detects lifecycle scripts that execute arbitrary commands during package installation (preinstall, postinstall, install hooks in package.json, setup.py with cmdclass, etc.).

Legitimate packages rarely need install scripts. Malicious packages use them to exfiltrate environment variables or download second-stage payloads.

Detects references to native binary compilation in manifests (.gyp files, node-gyp dependency, gypfile: true, native extensions in setup.py).

Detects package names that could be substituted via a private registry attack — internal-looking names (single word, no scope, matching known internal naming patterns) published to the public registry.

Detects unusual metadata in package manifests: empty or missing descriptions, mismatched author fields, extremely short package names, version 0.0.0, or missing repository fields.

Detects indicators of intentional obfuscation in package config: base64-encoded inline scripts, hex-encoded strings, eval-based execution patterns, or minified code embedded directly in package.json.

Detects declared dependencies on libraries with known unsafe deserialization behavior (serialize-javascript, node-serialize, PyYAML with unsafe loader, Java's ObjectInputStream wrappers).

Detects dependency or config patterns that enable dynamic code execution (e.g. vm2, eval packages, template engines with dangerous defaults declared as runtime deps).

Detects packages or scripts that extract archives (.tar, .zip, .tgz) without path traversal protection, enabling zip-slip attacks.

Detects manifest patterns that imply unbounded memory or CPU usage: packages that parse untrusted XML/HTML without limits, regex engines without timeout support, streaming parsers without backpressure.

Detects API keys, tokens, or credentials embedded directly in manifest files (package.json config blocks, .npmrc files, pyproject.toml tool config sections).