返回资讯中心
外部精选
软件工程
#开发者实时流

RepoTrust 推出代码库可信度评分工具,可识别 AI 生成代码占比

来源标题:RepoTrust – Determine if a repo is trustworthy enough to use

RepoTrust 上线公开仓库可信度评估服务,可在数秒内给出代码构建方式、工程质量与安全性的证据化评分。该工具将 AI 代码溯源、代码质量和安全分开打分,每项评分可追溯到具体提交和代码行,并标注 AI 生成代码占比及其置信等级。评估仅做静态分析,在无网络的隔离容器中运行,不安装、构建或执行代码。

Hacker News 最新harryhoffman2026/11/14 · 3 天前4 分钟阅读

以下正文由 AI 翻译自 Hacker News 最新(原文版权归原站所有),可能存在偏差,重要内容请以原文为准。

Stars and a polished README no longer tell you whether code was reviewed, tested or secured, or how much of it an AI wrote. RepoTrust gives you an evidence-backed read on any public repository in seconds: how it was built, how well it's engineered, and whether it's safe to depend on.

Check a repo Not on the board yet? We'll take you to the submit form, and it's usually scored within the hour.

AI provenance, code quality and security are separate questions. One number would hide the answer you need.

Each score links to the commits, files and lines behind it, so you can check our work in a minute.

Static analysis only, in isolated containers with no network. Nothing is installed, built or executed.

Measured against hundreds of labelled repositories, with accuracy published. Disputes fix the method for everyone.

The estimated share of the code that was AI-generated, with a label for how sure we are: explicit (the commits say so), likely, possible (weaker signals, such as AI tool configuration or LLM-style comments) or none detected. AI use is a fact to disclose, not a defect: AI-written code can score well on quality and security.

0–100 from complexity, duplication, lint density, tests and CI, documentation, type safety and review practice. Strong 80+, adequate 60–79, weak 40–59, poor below 40.

0–100 from secrets in code and history, static analysis, known-vulnerable dependencies, CI workflow risks and security posture. Same bands as quality. Flags such as suspicious-distribution warn of repos that look like malware lures.

Scores are automated signals with stated confidence and coverage, not verdicts. Read the evidence on each repository's page, and see the methodology for the details.

正文由 探How 从来源站点 RSS 同步,内容未经改写;遇到排版缺失或需要图片、视频时请以原文为准。

打开原文
去社区讨论 →