AI Security Review
scanned 5d ago · by lpm-firewall-aiLPM treats this as warn-only first-party agent extension lifecycle risk. No confirmed malicious install-time behavior. The package is a security scanning CLI/MCP server with user-invoked agent config, git hook setup, scanner downloads, and project report writes.
Decision evidence
public snapshot- dist/agent-init-VWOQP2MQ.js can create or append AGENTS.md, .cursor/rules/auditx.mdc, .cursorrules, .github/copilot-instructions.md, and .claude/skills/auditx/SKILL.md.
- dist/hook-W4INEOTH.js can install pre-commit/pre-push/post-merge/post-checkout hooks that run auditx on git operations.
- dist/chunk-PGZJFI55.js downloads scanner binaries from GitHub releases into ~/.auditx/bin and may run pip install semgrep.
- dist/auditx.js user command 'install' globally installs npm scanner packages and pip lizard.
- package.json has no preinstall/install/postinstall lifecycle hook.
- Agent config and git hook writes are activated by explicit CLI subcommands, not import or npm install.
- Network downloads are for named security scanners from GitHub release URLs and are package-aligned.
- No credential harvesting or exfiltration endpoint found; AI API keys are read only when --ai is requested.
- MCP server exposes an audit_codebase tool over stdio and does not register itself globally.
Source & flagged code
5 flagged · loading sourceThis package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
dist/auditx.jsView on unpkgPackage source invokes a package manager install command at runtime.
dist/auditx.jsView on unpkg · L838Package source references weak cryptographic algorithms.
dist/auditx.jsView on unpkg · L19