AI Security Review
scanned 2h ago · by lpm-firewall-aiExplicit runner and tunnel commands create an outbound control channel to the configured API. A runner can be instructed to probe network-reachable URLs, and a tunnel can relay local HTTP responses.
Decision evidence
public snapshot- `lib/runner.mjs` polls the configured API for jobs and fetches server-supplied internal target URLs.
- Runner posts target metadata, hostname, titles, server headers, and broken-link results back to the API.
- `lib/tunnel.mjs` relays local HTTP requests and full response bodies through the configured control plane.
- `bin/aegis.mjs` installs a Git pre-push hook only after explicit `aegis hooks install`.
- The `dev` command executes only a command supplied after `--`; Windows shell use is platform-specific.
- `package.json` has no preinstall, install, postinstall, or prepare lifecycle scripts.
- The package contains only readable JavaScript source; no bundled binaries, bytecode, or staged payloads.
- No credential harvesting beyond explicitly supplied `AEGIS_TOKEN` and scan password inputs was found.
- No eval, VM, native loading, hidden persistence, or import-time network action was found.
- Hook mutation is opt-in, marker-guarded, and restores a backed-up hook on uninstall.
- Default API traffic targets the package-aligned `https://app.aegisrunner.com/api/v1` endpoint.
Source & flagged code
4 flagged · loading sourceThis package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
bin/aegis.mjsView on unpkgSource combines command execution, command-output handling, and outbound requests; review data flow before blocking.
bin/aegis.mjsView on unpkg · L517Source gates dangerous network, credential, or execution behavior behind CI, host, platform, time, or geo fingerprint checks.
bin/aegis.mjsView on unpkg · L9