Skip to content
Back to skills

Priority Sensitivity Testing

ASecurity

Tactic: 扰动评分权重,检验 gap 排序对权重选择的稳健性

  • 417 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added May 27, 2026
researchgotesting

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned May 27, 2026

npx -y skills add yogsoth-ai/de-anthropocentric-research-engine --skill priority-sensitivity-testing --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Priority Sensitivity Testing?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Priority Sensitivity Testing
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/yogsoth-ai-priority-sensitivity-testing/badge)](https://www.skillsdirectory.com/skills/yogsoth-ai-priority-sensitivity-testing)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: priority-sensitivity-testing
description: "Tactic: 扰动评分权重,检验 gap 排序对权重选择的稳健性"
version: 1.0.0
category: hypothesis-formation
type: tactic
campaign: gap-prioritization
sops:
  - ahp-weighting
  - weight-perturbation
  - priority-synthesis
dependencies:
  skills:
    - subagent-spawning
---

# Priority Sensitivity Testing

对已有的 gap 评分矩阵施加权重扰动,检验最终排序是否对权重选择敏感,从而判断优先级决策的可信度。

## 编排意图

评分矩阵的最终排序依赖于各维度的权重设置。若权重稍有变化排序就大幅改变,则决策不可靠;若排序在合理权重范围内保持稳定,则优先级结论更有说服力。

本 tactic 先建立基线权重(AHP 或等权重),再系统性地扰动权重(±20%),观察排序变化,最终给出稳定性判定。

## 可用 SOPs

| SOP | 职责 | 何时调用 |
|-----|------|---------|
| ahp-weighting | 使用 AHP 方法从维度重要性判断中推导权重向量 | 第一步,建立基线权重 |
| weight-perturbation | 对每个维度权重施加 ±20% 扰动,重新计算排序 | 第二步,系统性扰动 |
| priority-synthesis | 汇总所有扰动场景的排序结果,合成稳定性报告 | 第三步,合成结论 |

## 编排模式

**Default(标准流程)**
1. ahp-weighting:从用户/策略提供的维度重要性判断推导基线权重;若无判断则使用等权重
2. weight-perturbation:对每个维度依次施加 +20% 和 -20% 扰动(其余维度等比例调整保持总和为 1),产出每个扰动场景的排序
3. priority-synthesis:统计每个 gap 在所有扰动场景中的排名分布,输出稳定性判定

**Simplified(S tier 或快速模式)**
- 跳过 ahp-weighting,直接使用等权重作为基线
- 仅扰动最高权重维度(±20%),产出 2 个扰动场景
- priority-synthesis 给出简化稳定性报告

**Deep(L tier 或高风险决策)**
- ahp-weighting 使用完整 AHP 成对比较(维度间两两比较)
- weight-perturbation 扰动幅度扩展至 ±30%,并增加极端场景(某维度权重设为 0)
- priority-synthesis 额外输出 Spearman 秩相关系数(基线排序 vs 每个扰动排序)

## Minimum Yield

- 基线权重向量(所有维度权重之和 = 1)
- 至少 3 个扰动场景的排序结果(每个场景标注扰动内容)
- 每个 gap 的排名稳定性统计(最高排名、最低排名、众数排名)
- 最终稳定性判定:**Stable**(所有场景前 N 名不变)/ **Partially Sensitive**(前 N 名有 1-2 个变化)/ **Highly Sensitive**(前 N 名变化超过 2 个)

## Yield Report

执行结束后向调用方 strategy 报告:
- 基线权重来源(AHP 推导 / 等权重 / 用户指定)
- 扰动场景数量
- 稳定性判定结论
- 最不稳定的 gap(排名波动最大)及建议(是否需要补充证据再决策)

Files in this skill

  • SKILL.md2.9 KB
  • prompt.md3.8 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.