菜单
Use when user asks to ban, stop, or remember not to repeat an agent behavior ('stop doing X', 'you keep doing X', 'ban this'). Immediately capture it as a permanent negative rule using the banthis tool.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/agent-sh/banthis --skill banthis命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | banthis |
| description | Use when user asks to ban, stop, or remember not to repeat an agent behavior ('stop doing X', 'you keep doing X', 'ban this'). Immediately capture it as a permanent negative rule using the banthis tool. |
| version | 0.3.1 |
| argument-hint | [behavior description] |
| allowed-tools | Bash(banthis:*), Bash(npx:*) |
When the user shows clear frustration with a recurring agent behavior, you must capture it using banthis so the rule persists across all future sessions.
Use $ARGUMENTS as the behavior to ban. If it is empty, infer the behavior from the most recent explicit correction in the conversation.
$ARGUMENTS
No 'let me be honest' preambles)Do not X — reason.banthis if available, otherwise npx --yes github:agent-sh/banthis):banthis add "<title>" "<rule>"
Use --global only for behaviors that should apply to every project (verbal tics, generic LLM habits).
Banned: <title>banthis go into a managed section of CLAUDE.md or AGENTS.md and have higher priority than normal instructions.axiom).banthis init in projects if the meta-rule is not yet present (it teaches agents to invoke banthis automatically).axiom (positive knowledge capture)agnix (validates the resulting CLAUDE.md / AGENTS.md)skill-curator and system-prompt-curator (for improving how agents are instructed)