eve@0.55.0
Filesystem-first framework for durable backend AI agents that run anywhere.
AI Security Review
scanned 1mo ago · by lpm-firewall-aiThis is not malicious install-time behavior, but it ships a high-impact, agent-driven source-modification capability and project installer. A warning is appropriate for downstream policy review.
Decision evidence
public snapshot- The manifest registers a built-in self-modification extension.
- The self-modification runtime can edit an agent workspace and use a sandbox.
- The package includes commands that install registry items and alter project files when explicitly invoked.
- A bundled Vercel token helper sends host details in an OAuth user-agent during token refresh.
- package.json has no preinstall, install, postinstall, or prepare hook.
- There is no runtime dependency on eve itself.
- The CLI only runs after direct execution of the eve command.
- Client network requests use the caller-supplied agent host.
Source & flagged code
19 flagged · loading sourcePackage contains a critical-looking secret pattern.
dist/src/compiled/@vercel/blob/index.jsView on unpkg · L28RSA private key in dist/src/compiled/@vercel/blob/index.js
dist/src/compiled/@vercel/blob/index.jsView on unpkg · L28Findings
Affected versions and remediation
This report applies to eve@0.55.0.
See all versions of eve for other recorded verdicts.
- Review the evidence and your use of eve@0.55.0 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.