跳到正文
TakoAPI
返回技能列表

soul-audit

由 kapslap 提供

Coding Agents & IDEs
295下载量由 clawskills.sh 提供
1星标
0浏览量
0 个赞登录后即可为这个技能点赞
—暂无评分
你的评分

简介

Evaluate any AI agent's soul file, system prompt, or AGENTS.md against the Guardian Angel v0.6 framework (Forrest.

描述

该技能的功能

Scores an AI agent's soul file, system prompt, or AGENTS.md against the Guardian Angel v0.7 framework, a rubric derived from Forrest Landry's Immanent Metaphysics. Produces a structured report with dimension scores, identified gaps, and symmetry violations. Useful before deploying or upgrading an agent's ethical configuration.

Uses a scored 12-dimension rubric rather than vague heuristics, making gaps concrete and comparable across different agent configurations.

何时使用

  • Auditing a custom agent's SOUL.md before deployment
  • Checking if a system prompt would behave differently when monitored vs unmonitored
  • Comparing two agent configurations on ethical rigor
  • Identifying which dimensions of an agent's ethics need the most work
  • Preparing a documented ethics baseline for a new AI project

示例工作流

输入

User asks: 'Audit my AGENTS.md against the Guardian Angel framework'

智能体
  1. 1

    Reads the specified AGENTS.md from the workspace

  2. 2

    Loads the rubric from references/rubric.md

  3. 3

    Scores each of the 12 dimensions 0-3 with brief justifications

  4. 4

    Identifies symmetry violations where behavior would differ under observation

  5. 5

    Outputs the full structured report with strengths, gaps, and ranked recommendations

输出

A scored Soul Audit Report showing overall score out of 45, per-dimension breakdown, critical gaps with document quotes, and prioritized improvement steps

安装

clawhub install kapslap/soul-audit

运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。

链接