KingOfZhao Decision Framework

决策框架 Skill —— 已知/未知驱动的结构化决策系统，置信度加权+红线拦截+决策记忆

Security Scan

This is primarily an instruction-only, documentation-style decision-framework that requests no credentials, but it claims an install command and a Python API that are not present in the package—an inconsistency you should resolve before trusting it.

Purpose & Capability

The name/description (decision framework for known/unknown-driven decisions) matches the SKILL.md content (KUWR method, scoring, redline checks, decision memory). However the SKILL.md shows a Python usage example (from skills.decision_framework import DecisionFramework) and an install command (clawhub install decision-framework) while the skill bundle contains no code files or install spec. That mismatch (documentation claiming code/API that isn't included) is unexplained.

Instruction Scope

The runtime instructions are limited to decision structuring, scoring, redline checks, and recording decision-memory files (decisions/{date}_{id}.md). They do not instruct reading unrelated system files or exfiltrating secrets. The decision-memory convention implies writing files to the workspace; that is expected for this purpose but should be noted.

Install Mechanism

No install spec is present (instruction-only), which is low risk. The SKILL.md nevertheless references an external installer command (clawhub install decision-framework). Because no install artifacts or code are included, it's unclear whether that command fetches trusted code or whether the platform is expected to provide the implementation—this ambiguity should be resolved.

Credentials

The skill requests no environment variables, no credentials, and no config paths. The declared requirements are minimal and proportionate to a documentation/methodology skill.

Persistence & Privilege

always is false and there's no indication the skill requests elevated or persistent platform privileges. It includes instructions to save decision records to a workspace path, which is reasonable for a memory/recording feature.

What to consider before installing

This appears to be a documentation-only decision framework (method + examples) rather than an actual packaged Python module. Before installing or running anything: 1) Confirm the source—visit the provided GitHub link and verify whether implementation files exist and are trustworthy. 2) Ask the author or maintainer where the referenced Python module and the 'clawhub install' artifacts come from; the package bundle you provided contains no code. 3) If you accept running this, run it in an isolated workspace because the framework expects to write decision records (decisions/{date}_{id}.md) which may contain sensitive decision rationale. 4) Because no credentials are requested, there is no obvious secret-exfiltration risk in the files shown, but the missing implementation and ambiguous install path are the primary concerns to resolve. If the author supplies the missing code, re-evaluate that code (or share it) before trusting automated execution.

Decision Framework Skill

元数据

字段	值
名称	decision-framework
版本	1.0.0
作者	KingOfZhao
发布日期	2026-03-31
置信度	96%

来源碰撞

self-evolution-cognition (已知/未知分离)
        ⊗
entrepreneur-cognition (商业决策)
        ⊗
ai-growth-engine (度量化)
        ↓
decision-framework (通用决策系统)

核心哲学

好决策 ≠ 好结果。好决策 = 在已知信息下，用结构化方法选择期望值最高的选项。 结果可能受运气影响，但决策过程可以优化。

四步决策法（KUWR）

K — Known 已知:    列出所有已知事实，标注置信度
U — Unknown 未知:  列出所有未知项，评估影响和获取成本
W — Weigh 权衡:    对每个选项做已知/未知加权评分
R — Record 记录:   记录决策+理由+预期结果，事后对比

决策评分公式

Option Score = (Known_Benefit × confidence) - (Known_Cost × confidence) 
               + (Unknown_Upside × probability × 0.5) 
               - (Unknown_Downside × probability × 1.5)

注意: 未知下行风险权重(1.5) > 未知上行收益权重(0.5)
→ 这是"不对称风险偏好": 优先避免灾难性错误

红线拦截

决策前自动检查红线:
  🔴 此决策是否触碰职业红线？→ 如果是，自动拦截
  🔴 此决策是否不可逆？→ 如果是，强制等待24h冷静期
  🔴 此决策是否影响他人？→ 如果是，强制咨询受影响方
  🔴 已知信息置信度是否 <50%？→ 如果是，建议先补充信息

决策记忆

decisions/{date}_{id}.md
  - 问题描述
  - 已知/未知列表
  - 各选项评分
  - 最终选择 + 理由
  - 预期结果
  - [事后] 实际结果 + 偏差分析
  - [事后] 教训（更新决策参数）

安装命令

clawhub install decision-framework

调用方式

from skills.decision_framework import DecisionFramework

df = DecisionFramework(workspace=".")

result = df.decide(
    problem="是否应该切换到新的技术栈？",
    options=["保持现状", "渐进迁移", "完全重写"],
    known=["当前技术栈稳定运行2年", "新栈性能提升40%", "团队3人熟悉新栈"],
    unknown=["迁移周期", "业务中断风险", "学习曲线"],
    redlines=["不中断生产服务"]
)
print(result.rankings)       # 选项排名 + 评分
print(result.redline_check)  # 红线检查结果
print(result.recommendation) # 建议选择 + 理由

学术参考

1. A Survey of Self-Evolving Agents — 决策自进化
2. SAGE: Multi-Agent Self-Evolution — 多Agent决策
3. Memory in the Age of AI Agents — 决策记忆