No red flags found in any of the 11 categories — no credential harvesting, no data exfiltration, no curl-pipe-shell installer. Scanned against the SlowMist agent-security taxonomy, refreshed every 8 hours. Full audit →
by jev-chat · LLM Plugin · ★ 453
Last updated: · Indexed by AgentSkillsHub · Auto-synced every 8h
🔒 Is jev-chat-jarvis-mac safe to install? View the security audit →
jev-jarvis（macOS） 微信弹出一条消息 → 悬浮窗立刻告诉你这句话的真实意图、风险几级、该怎么回。 Mac 版 V0：纯只读、零封号风险——不注入、不 hook、不解密数据库，只是「看屏幕 + 本地模型判断」。 交流反馈 用着有问题、想提需求、或者想一起改，扫码进群（二维码 7 天失效，过期了在 issue 说一声）： 现在能做什么 意图 + 风险：8 类意图零样本 86.4%（22 条实测），风险 0–9 分级 + 行动建议，本地模型一次前向出全分布 候选回复：按选中的「话术」生成（内置 10 种、最多 3 种并发、keep-alive 复用连接）→ 先上屏（流式端点逐行、其余整包，标「排序中」）→ 本地模型排序后原位重排，压掉踩雷选项 YOLO 检测框（可选）：把每次 OCR 命中的消息实时框在微信窗口上——对方/我/发送者分色 + 真实置信度，被判断那条带「意图·风险」标签；点击穿透、不进抓图。env 里 启动即开，菜单栏 J 图标随时切换 快：消息一出现预判+生成同时起跑（latest-wins，刷屏中间态自动作废），停稳窗口把生成耗时整个藏进去 防刷屏：消息停稳才分析（上限 1.2 s，连续 3 跳安静最早 0.7 s 开闸）、两次分析最小间隔 2 s；分析期间读屏不中断 悬浮窗跟随微信窗口（非激活不抢焦点、多显示器不跳屏）；「复制」/「填入」走辅助功能接口，输入框已有字时追加不覆盖，写完读回确认 用法 只想用：Releases 下载 ，解压拖进「应用程序」。第一次打开要右键 → 打开（没做公证，双击会被 Gatekeeper 拦）；首次启动联网装依赖约 3 s，授予「屏幕录制」权限后退出重开。 从源码跑（微信在运行、终端已授予「屏幕录制」）：。「填入」另需「辅助功能」权限，第一次点会弹系统授权框。分层自测： 面板读法 微信浅色配色（深色模式下也保持浅色），从上到下： 等待微信消息… ← 状态行 这个方案今天能给我吗？ ← 被判断的消息 来自 王总 · 上文：昨天的评审意见…
| Stars | 453 |
| Forks | 130 |
| Language | Python |
| Category | LLM Plugin |
| License | MIT |
| Quality Score | 62.0004481755788/100 |
| Open Issues | 19 |
| Last Updated | 2026-09-30 |
| Created | 2026-09-21 |
| Platforms | python |
| Est. Tokens | ~15k |
These tools work well together with jev-chat-jarvis-mac for enhanced workflows:
Looking for a jev-chat-jarvis-mac alternative? If you're comparing jev-chat-jarvis-mac with other llm plugin tools, these 6 projects are the closest alternatives on Agent Skills Hub — ranked by topic overlap, star count, and community traction.
AI gets the context. Not your private data. Local-first privacy proxy for browser chat, AI APIs, and coding ag
Local-first personal agentic OS and everything app for coding, knowledge work, web design, automations, and ar
MCP server that lets Claude Code and other AI agents read and search large PDFs, one file or a whole folder: a
AI as Workspace - An elegant AI chat client. Full-featured, lightweight. Support multiple workspaces, plugin s
Compile a demonstrated GUI workflow into a deterministic, locally executable program. Zero model calls on heal
High-performance OpenAI and Anthropic compatible LLM inference server for Apple Silicon. Native MLX, continuou
Explore other popular llm plugin tools:
jev-chat-jarvis-mac is 聊天悬浮窗助手（macOS）：屏幕感知 + 本地小模型判断意图与风险，按话术生成回复候选。纯只读。. It is categorized as a LLM Plugin with 453 GitHub stars.
jev-chat-jarvis-mac is primarily written in Python. It covers topics such as llm, local-first, macos.
You can find installation instructions and usage details in the jev-chat-jarvis-mac GitHub repository at github.com/jev-chat/jev-chat-jarvis-mac. The project has 453 stars and 130 forks, indicating an active community.
jev-chat-jarvis-mac is released under the MIT license, making it free to use and modify according to the license terms.
The top alternatives to jev-chat-jarvis-mac on Agent Skills Hub include pasteguard, CoWork-OS, pdf-mcp. Each offers a different approach to the same problem space — compare them side-by-side by stars, quality score, and community activity.
Grades come from a rule-based scan built on the SlowMist agent-security taxonomy, covering 11 red-flag categories including credential harvesting, data exfiltration, and curl | sh installers. It is a first-layer scan, not a manual audit — we say so rather than overstate it.
The scale of the problem is documented independently: Liu et al. (2026), in a study of 31,132 agent skills, report that 26.1% contain security vulnerabilities. Our own full-catalog census is published as a citable open dataset.
Sources & who's responsible: