Evidence for AI clean decision
- The manifest declares build and development scripts but no install lifecycle script.
- The setup command documents the Pushary MCP endpoint and its PUSHARY_API_KEY authentication for connecting agent clients.
- The setup flow reads a Pushary API key from explicit input, PUSHARY_API_KEY, or saved configuration for use with Pushary.
- The shell-profile alias writer is an ordinary named helper that appends a Pushary alias; it is not called at module top level in the inspected excerpt.
- The Claude wrapper sends the API key to the Pushary command-drain endpoint as part of its agent-control feature.
Evidence against
- The package exposes powerful setup, agent bridge, and shell-profile modification capabilities, but the reviewed actions serve its advertised human-in-the-loop agent integration.
- No credential flow to an unrelated host or install-time data collection was identified in the inspected paths.
Behavioral surface
SourceChild process · Crypto · Dynamic require · Environment variables · Filesystem · Network · Shell · Web socket
Supply chainHigh entropy strings · URL strings
ManifestNo manifest risk signals triggered.
Scanned 190 files · 3.19 MB of source
9 external domains referenced
- claude.ai
- discord.gg
- fx.sh
- nodejs.org
- pushary.com
- registry.npmjs.org
- schemas.microsoft.com
- www.apple.com
- www.pushary.com