claude-anti-ban
wujiajun4/claude-anti-ban
Claude anti-ban assistant — 5-dimension self-check + 4-red-zone detection + ban-fallback SOP. Self-built skill for Claude Code, integrates ~/Desktop/claude 防止被封指南/ knowledge.
Instalação
npx convez-ai wujiajun4/claude-anti-ban1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
claude-anti-ban
Claude anti-ban assistant — 5-dimension self-check + 4-red-zone detection + ban-fallback SOP. 整合 Claude 5 维度风控 (IP / 支付 / 设备指纹 / 行为 / 时间) + 双轨制架构 + 应急 SOP. 下次 Claude 出问题 (被封 / 弹 CAPTCHA / 401 / IP 黑) → 调本 skill → 30s 定位 + 30min 恢复.
English
What it does
claude-anti-ban is a skill (not a tool) that triggers when you mention:
- "Claude 被封" / "I got banned" / "弹 CAPTCHA" / "CAPTCHA popped"
- "401" / "IP 黑" / "fingerprint" / "risk check"
- "5 维度" / "5-dimension check" / "风控" / "风控警告"
Once triggered, the agent (running on Claude Code) walks you through the 5-dimension self-check, identifies which red zone you hit, and (if needed) runs the ban-fallback SOP (data backup → switch to MiniMax → wait 6-12h before appealing).
Quick reference
| Section | Purpose |
|---|---|
| §1 5 dimensions | IP / payment / fingerprint / behavior / time — self-check |
| §2 4 red zones | 5 IP/h / shared account / >40 msg/min / multi-device |
| §3 Dual-track | Pro Desktop (risk side) + MiniMax Claude Code (independent) |
| §4 Ban-fallback SOP | 1h action + 1-7d scenario + 1-3m long-term |
| §5 Detection | 1-line curl to verify "really banned" vs other issues |
| §6 Your environment | Template — fill in your IP / timezone / Pro status |
| §7 Fingerprint 6 | Run --dry-run to see your 6 fingerprints |
| §8 Subscription | 3 methods (WildAI / Apple gift card / Google Pay) |
Install
This is a self-built skill in ~/.claude/skills/claude-anti-ban/. After
clone, just put it in your skills directory:
# Option A: clone into your skills folder
git clone https://github.com/wujiajun4/claude-anti-ban ~/.claude/skills/claude-anti-ban
# Option B: symlink (for development)
ln -s /path/to/claude-anti-ban ~/.claude/skills/claude-anti-ban
The agent will auto-load it on next session.
License
MIT — see LICENSE.
中文
做什么
claude-anti-ban 是个 skill (不是工具), 触发词:
- "Claude 被封" / "弹 CAPTCHA" / "401" / "IP 黑" / "设备指纹" / "risk check"
- "5 维度" / "风控" / "风控警告" / "自查" / "防封"
触发后, agent (Claude Code) 带你走 5 维度自查, 识别踩了哪个雷区, 出问题 走 ban-fallback SOP (备份 → 切 MiniMax → 等 6-12h 再申诉).
8 段速查
| § | 内容 |
|---|---|
| §1 5 维度 | IP / 支付 / 设备指纹 / 行为 / 时间 — 出问题先查这 5 条 |
| §2 4 雷区 | 5 IP/h / 共享账号 / >40 msg/min / 多设备同时 |
| §3 双轨制 | Pro Desktop (风控端) + MiniMax Claude Code (独立) |
| §4 ban-fallback SOP | 第 1 小时 / 第 1-7 天 / 长期 |
| §5 检测 | 1 行 curl 验真封 vs 别的原因 |
| §6 你的环境 | 模板 — 填你自己的 IP / 时区 / Pro 状态 |
| §7 Fingerprint 6 | 跑 --dry-run 看你的 6 指纹 |
| §8 充值 | 3 种方法 (WildAI / Apple 礼品卡 / Google Pay) |
安装
自建 skill, 装在 ~/.claude/skills/claude-anti-ban/. 装法:
# 方式 A: clone 到 skills 目录
git clone https://github.com/wujiajun4/claude-anti-ban ~/.claude/skills/claude-anti-ban
# 方式 B: 软链接 (开发时)
ln -s /path/to/claude-anti-ban ~/.claude/skills/claude-anti-ban
下次开 Claude Code 自动加载。
License
MIT — 见 LICENSE。
Changelog
See SKILL.md changelog block — newest first, includes v0.1.0 initial release.
Author
wujiajun4