clavue-v1@1.18.1
Clavue v1: lightweight four-slot (main/light/plan/review) multi-model coding CLI on the execution-first engine — overlapped cross-family review, official Clavue cloud preset (api.clavue.com) alongside any Messages-compatible provider, learned context ceil
AI Security Review
scanned 4d ago · by lpm-firewall-aiSource inspection found no install lifecycle hook or concrete attack behavior. The flagged dynamic evaluation, process execution, and network code are used in serializer, CLI, provider, and proxy functionality; inspected account requests target the package's own service.
Decision evidence
public snapshot- package.json exposes a CLI and contains no install lifecycle hook.
- The provider router configures Clavue account requests to api.clavue.com and www.clavue.com.
- The flag client sends authentication headers to the Clavue API after a trust check or in a non-interactive session.
- The OpenAI adapter handles request URLs and headers as part of its provider client.
- The command runner spawns the requested command with caller-supplied arguments and options.
- AJV generates serializer code from its source template; the sandbox HTTP proxy handles CONNECT requests.
- The bundled code includes dynamic code generation and child-process execution, but the inspected contexts show serializer generation and a command runner rather than an attack.
- The runtime includes network and credential handling, with inspected account requests directed to the package's Clavue service.
Source & flagged code
12 flagged · loading sourcePackage source references child process execution.
dist/mao-command.jsView on unpkg · L4Findings
Affected versions and remediation
This report applies to clavue-v1@1.18.1.
See all versions of clavue-v1 for other recorded verdicts.
- Review the evidence and your use of clavue-v1@1.18.1 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.