← Back to Skills Marketplace
93
Downloads
1
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install deep-loop-thinker
Description
多轮深度思考技能。借鉴OpenMythos循环推理架构，每次循环都注入新输入，逐轮深化理解。适用于重要决策、复杂问题、创意生成、问题发现。
README (SKILL.md)
Deep Loop Thinker Skill v2.1
核心原理
借鉴OpenMythos的Recurrent-Depth Transformer：
- 同一层多次运行，每次有新输入注入
- 隐藏状态h_t更新：
h_{t+1} = A·h_t + B·e + Transformer(h_t, e) - 每次循环不是重复，而是递进
用户需求追踪
谁会使用这个技能？
- 创业者做重大决策
- 分析师处理复杂问题
- 开发者解决技术难题
- 研究者探索未知领域
- AI Agent增强推理能力
用户痛点
- 决策前思考不够全面
- 容易忽视风险
- 缺乏系统性反思
- 决策后不复盘
循环架构
问题输入
↓
[Prelude层] — 提取关键要素
↓
[Recurrent Block] — 多轮递进思考
↑____↓
↓ (每轮注入新洞察)
[Coda层] — 综合输出
↓
行动计划 + 自我反思
多轮思考模式
| 轮次 | 思考类型 | 核心问题 |
|---|---|---|
| 1 | 直觉捕捉 | 第一反应？情绪？ |
| 2 | 利益分析 | 谁受益？谁受损？ |
| 3 | 风险挖掘 | 最坏情况？3年后还重要吗？ |
| 4 | 本质洞察 | 根本原因？规律？ |
| 5 | 行动设计 | 第一步？Plan B？ |
| 6 | 反思验证 | 假设可靠？盲点在哪？ |
用户反馈收集
每次使用后记录：
【用户反馈】
- 用户是谁：{匿名/具体描述}
- 帮助程度：{1-5星}
- 改进建议：{用户的具体反馈}
- 下次需要：{用户的实际问题}
迭代记录
| 版本 | 日期 | 更新内容 |
|---|---|---|
| 1.0 | 2026-04-21 | 初始版本 |
| 2.0 | 2026-04-22 | 增加6轮思考 |
| 2.1 | 2026-04-22 | 增加用户追踪机制 |
质量标准
好的思考报告：
- ✅ 有具体数据/例子支撑
- ✅ 能回答"为什么"
- ✅ 第一步可立即执行
- ✅ 包含风险预案
- ✅ 诚实承认盲点
触发条件
- 问题影响超过3个月
- 涉及多方利益
- 情绪波动明显
- 反复纠结无法决定
- 问题反复出现
约束
- 简单问题不要过度思考（杀鸡不用牛刀）
- 不要超过6轮（防止无限循环）
- 保持诚实，不要自欺欺人
- 行动比完美计划重要
- 记录用户反馈用于迭代
Usage Guidance
This skill appears coherent and low-risk: it runs a local Python script that prints structured multi-round prompts and requires no network access or credentials. Before installing or giving it execution rights, you may want to: (1) run the Python script locally to verify behavior, (2) confirm the SKILL.md/README expectations about recording feedback (the shipped script does not persist logs), and (3) ensure you trust the source since the package contains a code file even though it is small and benign. If you plan to let an agent invoke skills autonomously, remember autonomous invocation is platform-default — consider limiting that capability if you do not want skills to run without explicit approvals.
Capability Analysis
Type: OpenClaw Skill
Name: deep-loop-thinker
Version: 2.1.0
The skill is a structured reasoning framework designed to help an AI agent perform multi-round deep thinking. The Python script (deep_loop_thinker.py) is a simple utility that prints conceptual prompts for different stages of analysis, and the documentation (SKILL.md) provides a logical roadmap for the agent to follow. There is no evidence of network activity, file system manipulation, data exfiltration, or malicious prompt injection.
Capability Assessment
Purpose & Capability
Name/description (multi-round reasoning) matches the SKILL.md and the included Python script: both implement prompts and a looped thinking workflow. No unrelated binaries, credentials, or external services are requested.
Instruction Scope
SKILL.md describes workflow, triggers, and a user-feedback recording format. The included Python script implements the interactive multi-round prompts and printing only. There is a minor inconsistency: SKILL.md states that user feedback is recorded for iteration, but the script does not persist feedback to disk or transmit it anywhere — it only prints. This is a functional mismatch but not a security concern.
Install Mechanism
No install spec is present (instruction-only). The only code file is a small local Python script that runs interactively; nothing is downloaded or executed from third-party URLs.
Credentials
The skill requires no environment variables, credentials, or config paths. The code does not access environment variables, secret names, or external services.
Persistence & Privilege
The skill is not always-enabled and uses normal agent invocation settings. It does not modify other skills or system-wide configs and does not persist data or install agents/services.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install deep-loop-thinker - After installation, invoke the skill by name or use
/deep-loop-thinker - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.1.0
增加用户追踪机制和反馈收集
v1.0.0
Initial release of deep-loop-thinker, a multi-turn deep thinking skill inspired by the OpenMythos recurrent reasoning framework.
- Supports multi-round reasoning with new input injected each iteration for deeper understanding.
- Provides structured templates for problem analysis, including intuition, stakeholder interests, risk, essence, action steps, and reflection.
- Tailors reasoning depth to problem complexity (1–6 rounds).
- Ideal for important decisions, complex issues, creative ideation, and root cause discovery.
- Encourages actionable insights, honesty about uncertainty, and risk mitigation in the output.
Metadata
Frequently Asked Questions
What is Deep Loop Thinker?
多轮深度思考技能。借鉴OpenMythos循环推理架构，每次循环都注入新输入，逐轮深化理解。适用于重要决策、复杂问题、创意生成、问题发现。 It is an AI Agent Skill for Claude Code / OpenClaw, with 93 downloads so far.
How do I install Deep Loop Thinker?
Run "/install deep-loop-thinker" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Deep Loop Thinker free?
Yes, Deep Loop Thinker is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Deep Loop Thinker support?
Deep Loop Thinker is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Deep Loop Thinker?
It is built and maintained by jaxint (@jaxint); the current version is v2.1.0.
More Skills