AI Security Review
scanned 2h ago · by lpm-firewall-aiLPM treats this as warn-only first-party agent extension lifecycle risk. The concrete risk is install-time execution that may install a first-party platform native binary package globally if the optional dependency is missing. No source evidence shows exfiltration, destructive behavior, foreign agent config mutation, or stealth persistence.
Decision evidence
public snapshot- package.json defines postinstall: node scripts/postinstall.js
- scripts/postinstall.js runs execSync npm install for @sciagent/cli-${platform}-${arch}@1.0.31 during install
- postinstall treats npm_lifecycle_event === postinstall as global context, so fallback install uses npm install -g
- bin/sciagent.js chmods and spawns a platform-specific native binary with user CLI args
- No credential, env, home directory, SSH, token, or secret harvesting found by source search
- No HTTP client or exfiltration endpoint in package source; only repository/docs URLs are present
- Platform package names are first-party @sciagent scoped optionalDependencies matching package version
- Runtime child_process use is the declared CLI wrapper launching the selected SciAgent binary
- No AI-agent control-surface writes, persistence setup, destructive file operations, eval, vm, or dynamic remote code loading in inspected JS
Source & flagged code
3 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgThis package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
scripts/postinstall.jsView on unpkg