AI Security Review
scanned 7d ago · by lpm-firewall-aiNo confirmed malicious attack surface is established. The package is a Dirac language interpreter/CLI with explicit user-invoked shell, eval, LLM, and agent features.
Decision evidence
public snapshot- dist/chunk-NUNYBLVJ.js exposes user-invoked <system> shell execution and <eval> AsyncFunction in the Dirac language runtime.
- dist/chunk-NUNYBLVJ.js and dist/chunk-PPH7KYKH.js can call LLM/embedding endpoints using env/config-selected providers.
- dist/agent-SVKTURXE.js starts a user-commanded background agent and writes under ~/.dirac.
- package.json has no preinstall/install/postinstall lifecycle scripts.
- dist/index.js import exports APIs without starting processes, network calls, or shell commands.
- dist/cli.js only runs when bin is invoked and reads explicit input/config files before executing Dirac code.
- Network use is package-aligned LLM/custom-local provider behavior, not hidden exfiltration.
- No credential harvesting loop, broad filesystem scan, remote payload download, or AI-agent control-surface mutation found.
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/chunk-NUNYBLVJ.jsView on unpkgPackage source references child process execution.
dist/chunk-NUNYBLVJ.jsView on unpkg · L1716Package source references dynamic require/import behavior.
dist/chunk-NUNYBLVJ.jsView on unpkg · L299Source spawns a local helper that also contains network and dynamic execution context; review data flow before blocking.
dist/shell-SUNLGPOM.jsView on unpkg · L20