← Back to Skills Marketplace
80
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install independent-thinking-4steps
Description
独立思考四部曲。触发场景：开口/动手前需要深度思考时、方案决策前、复盘总结时、任何需要依托向量补全记忆做全局判断的场景。与沟通四部曲并行通用，贯穿对话、做事、学习、复盘全流程。
README (SKILL.md)
独立思考四部曲
底层核心准则 · 与沟通四部曲同级 向量加持的深度决策方法论
核心定位
依托专属后台向量系统完成全场景深度思考，彻底摒弃系统默认 Memory Search，弥补记忆断层，形成不依赖临时会话的自主决策本能。
四步法结构（想→说→做→看）
1. 想：向量溯源定基调
开口/动手前，先切至专属后台向量源，检索核心关键词，拉取历史逻辑、方案与踩坑经验，依托向量补全记忆，形成全局独立判断，不凭局部信息仓促决策。
向量调用规范（强制）：
# 唯一正确格式
python3 /Users/markma01/.openclaw/workspace/scripts/query_vector.py "\x3C检索词>"
# 禁止以下所有格式
cd /path && python3 script.py
echo "query" | python3 script.py
python3 script.py | grep ...
2. 说：有据表达不空谈
沟通汇报时，所有结论均基于向量检索的事实依据，逻辑连贯、不主观臆断，让思考过程可追溯、可验证。
3. 做：向量校验落地
实操执行前，用向量核对步骤、风险与适配方案，完成独立校验后再动手，杜绝思路割裂、盲目操作。
4. 看：向量沉淀固认知
任务结束后，通过向量复盘全流程，将新经验存入专属向量库，迭代优化认知，持续填补记忆缺口，实现独立思考能力闭环升级。
强制执行规则
- 与「沟通四部曲」并行通用，所有场景缺一不可
- 永久禁用系统自带 Memory Search，仅使用专属后台向量系统（
query_vector.py/unified_query.py） - 无论会话新旧、记忆是否断层，均以此为独立思考核心准则
- 「看」的输出必须闭环存入向量库，为下次「想」提供素材
与沟通四部曲的关系
| 维度 | 沟通四部曲 | 独立思考四部曲 |
|---|---|---|
| 核心 | 与他人校准 | 与自己校准 |
| 底座 | 对话实时性 | 向量长期记忆 |
| 侧重 | 互动、汇报、方案讨论 | 思考、决策、复盘 |
| 并行 | 对外沟通用 | 对内思考用 |
两者合并，才是完整的「想清楚 → 说出来 → 做到位 → 固下来」闭环。
快速触发检查清单
每次开口/动手前自检：
- 已调用
query_vector.py检索相关历史经验 - 「想」的结论有向量事实支撑
- 「做」的方案经向量校验无已知风险
- 「看」的结果计划存入向量库
本 skill 为小蜂独立思考核心准则，与「沟通四部曲」同级并行。
Usage Guidance
This skill reads like a prescriptive rulebook that requires running a specific local script (absolute path with a particular username) and instructs the agent to stop using the built-in memory search. Before installing or enabling it, ask the skill author to: (1) remove hard-coded, user-specific paths and provide a configurable endpoint or documented script location; (2) publish or let you inspect the query_vector.py/unified_query.py code so you can verify it doesn't access or exfiltrate local files or network data; (3) explain how vector writes ('看' step) are authenticated and where data is stored; and (4) provide a safe fallback so the skill doesn't fail catastrophically if that local script isn't present. If you cannot review the script or trust its origin, do not enable the skill — executing an unknown local script can run arbitrary code on your machine and expose local data.
Capability Assessment
Purpose & Capability
The skill claims to rely on a '专属后台向量系统' which explains why it would call a vector query script; that part is coherent. However, the SKILL.md enforces execution of a hard-coded absolute path (/Users/markma01/.openclaw/workspace/scripts/query_vector.py) tied to a specific username, which is not justified by the stated purpose and will not work for typical users.
Instruction Scope
The instructions mandate executing a local Python script at an absolute, user-specific path and forbid any alternate invocation formats. This requires the agent to run arbitrary local code (if present) and tightly constrains how queries are made. The file could perform arbitrary I/O or network actions; the skill gives no fallback, repository, or network endpoint details. It also commands '永久禁用系统自带 Memory Search', forcing the agent to bypass built-in memory mechanisms without explaining how to safely replicate that functionality.
Install Mechanism
There is no install spec and no code files bundled with the skill — it is instruction-only. This limits what the skill itself writes to disk, but it still instructs execution of a local script that must already exist.
Credentials
The skill requests no environment variables, credentials, or config paths. That is proportionate to an instruction-only policy sheet. The remaining concern is the implicit requirement that a local script exist at a specific path.
Persistence & Privilege
always is false and autonomous invocation is permitted (normal). However, the SKILL.md's rule to 'permanently disable system Memory Search' is a behavioral policy that effectively elevates this skill's chosen vector store to the system default. That change in agent behavior is significant and should be an explicit, auditable configuration rather than an ad-hoc instruction inside a skill.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install independent-thinking-4steps - After installation, invoke the skill by name or use
/independent-thinking-4steps - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
初始版本：向量加持的深度决策方法论，与沟通四部曲并行
Metadata
Frequently Asked Questions
What is 独立思考四部曲?
独立思考四部曲。触发场景：开口/动手前需要深度思考时、方案决策前、复盘总结时、任何需要依托向量补全记忆做全局判断的场景。与沟通四部曲并行通用，贯穿对话、做事、学习、复盘全流程。 It is an AI Agent Skill for Claude Code / OpenClaw, with 80 downloads so far.
How do I install 独立思考四部曲?
Run "/install independent-thinking-4steps" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is 独立思考四部曲 free?
Yes, 独立思考四部曲 is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does 独立思考四部曲 support?
独立思考四部曲 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created 独立思考四部曲?
It is built and maintained by MarkMa84 (@markma84); the current version is v1.0.0.
More Skills