← 返回
未分类 中文

MIT Reasoning Lite

Provides a 3-step reasoning scaffold prompting agents to clarify objectives, check constraints, and articulate reasoning before significant decisions.
提供三步推理框架,在重要决策前引导代理明确目标、检查约束并阐述推理。
kgapol kgapol 来源
未分类 clawhub v1.0.0 1 版本 99731.9 Key: 无需
★ 0
Stars
📥 372
下载
💾 0
安装
1
版本
#latest

概述

MIT Reasoning Lite — Free Agent Decision Framework

Your AI agents are making decisions right now. Do you know how?

MIT Reasoning Lite gives your agents a 3-step reasoning scaffold that stops them from taking shortcuts you'll regret.

What This Skill Does

Forces structured thinking before action. Three steps, drop-in install, no configuration required.

The 3-Step Scaffold

Step 1 — Clarify the Objective

What is actually being asked? Not the surface request — the real intent. What outcome are we optimizing for?

Step 2 — Check Constraints

What rules, limits, history, or context applies to this decision? What's out of bounds?

Step 3 — Reason Before Acting

Articulate the reasoning before taking action. Show the work. If the reasoning doesn't hold up, the action doesn't happen.

When to Use

Add this to your agent's SKILL.md or AGENTS.md to activate for any significant decision:

  • Actions that affect external parties (sending emails, posting content, client communications)
  • Actions that are difficult to reverse (file changes, purchases, delegations)
  • Situations with ambiguous instructions
  • Any time an agent has had a "why did it do that?" moment before

Install

npx clawhub@latest install mit-reasoning-lite

Or manually: copy SKILL.md to your agent workspace.

Usage

Reference in AGENTS.md:

Before any significant decision or action, apply MIT Reasoning Lite:
1. Clarify the real objective (not just the surface request)
2. Check all constraints and rules that apply
3. State your reasoning before acting

Part of Trust AI Suite

MIT Reasoning Lite is the entry point to the Trust AI framework:

Free skills (ClawHub):

  • Accountable AI — governance & accountability
  • Proof of Work — verify agents completed what they claimed
  • MIT Reasoning Lite (this) — structured decision-making

Full version — MIT Reasoning ($19 on ClawMart):

  • 9-step full reasoning framework with branching logic
  • Decision logging with timestamped rationale
  • Uncertainty escalation (agent flags low-confidence decisions)
  • Cross-agent consistency checks
  • 6 pre-built decision templates

→ Full suite: Trust AI Stack on ClawMart

版本历史

共 1 个版本

  • v1.0.0 当前
    2026-05-07 04:54 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

ai-agent

self-improving agent

pskoett
捕获经验教训、错误及修正内容,以实现持续改进。适用于以下场景:(1)命令或操作意外失败;(2)用户纠正Claude(如“不,那不对……”“实际上……”);(3)用户请求的功能不存在;(4)外部API或工具出现故障;(5)Claude发现自身
★ 4,094 📥 820,772
ai-agent

Self-Improving + Proactive Agent

ivangdavila
自我反思+自我批评+自我学习+自组织记忆。智能体评估自身工作、发现错误并持续改进。
★ 1,390 📥 321,736
ai-agent

Skill Vetter

spclaudehome
AI智能体技能安全预审工具。安装ClawdHub、GitHub等来源技能前,检查风险信号、权限范围及可疑模式。
★ 1,233 📥 268,931