Evidence for warning
- Several distributed source files contain review-facing comments that tell a reviewer to allow a size finding and characterize the change as safe.
- The package manifest's prepublish hook runs cleanup, build, and shrinkwrap commands; the inspected source did not establish an install-time attack.
- The package manager starts subprocesses with the supplied command and arguments, optional working directory, and configured stdio.
Evidence against
- The review-facing comments have no demonstrated runtime effect, and the inspected package manager code handles command completion and errors.
- The proxy helper posts a caller-supplied token to a caller-supplied proxy URL; this does not establish hidden credential forwarding.
Behavioral surface
SourceChild process · Crypto · Dynamic require · Environment variables · Eval · Filesystem · Network · Shell
Supply chainHigh entropy strings · Minified · URL strings
ManifestNo manifest risk signals triggered.
Scanned 1,068 files · 6.92 MB of source
37 external domains referenced
- 127.0.0.1
- api.adaptorch.com
- api.anthropic.com
- api.commandcode.ai
- api.github.com
- api.githubcopilot.com
- api.kimi.com
- api.openai.com
- api.z.ai
- arxiv.org
- chatgpt.com
- claude.ai
- cli-chat-proxy.grok.com
- cli.github.com
- cloud.gitlab.com
- console.anthropic.com
- context7.com
- deepwiki.com
- example.com
- git-scm.com
- github.com
- gitlab.com
- learn.microsoft.com
- mariozechner.at
- mcp.context7.com
- mcp.exa.ai
- mcp.firecrawl.dev
- mise.jdx.dev
- mistral.ai
- omk.dev
- open.bigmodel.cn
- opencode.ai
- playwright.dev
- registry.npmjs.org
- terminal-browser.sh
- www.gamers.org
- www.w3.org