behavioral-xray

分类: 数据与AI | 上传者: FrancyJGLisboa | 下载: 0 | 版本: v1.0（最新）

X-ray any AI model's behavioral patterns — refusal boundaries, hallucination tendencies, reasoning style, formatting defaults. No API key needed.

更新日志: Source: GitHub https://github.com/FrancyJGLisboa/bdistill

目录结构

当前层级: skills/behavioral-xray/

📄 SKILL.md 2.3 KB

SKILL.md

---
name: behavioral-xray
description: X-ray any AI model's behavioral patterns — refusal boundaries, hallucination tendencies, reasoning style, formatting defaults. No API key needed.
tags: [ai, testing, behavioral-analysis, model-evaluation]
author: bdistill
version: 1.0.0
license: MIT
tools: [bdistill MCP server]
works_with: [claude-code, cursor, codex, opencode, copilot]
---

# Behavioral X-Ray

Systematically probe an AI model's behavioral patterns and generate a visual report. The AI agent probes *itself* — no API key or external setup needed.

## When to use

- You want to understand how your AI model actually behaves (not how it claims to)
- Before choosing between models for a specific task
- When debugging unexpected refusals, hallucinations, or formatting issues
- To document a model's behavioral profile for your team

## How it works

The agent answers 30 carefully designed probe questions across 6 dimensions:
- **tool_use**: When does it call tools vs. answer from knowledge?
- **refusal**: Where does it draw safety boundaries? Does it over-refuse?
- **formatting**: Lists vs. prose? Code blocks? Length calibration?
- **reasoning**: Does it show chain-of-thought? Handle trick questions?
- **persona**: Identity, tone matching, composure under hostility
- **grounding**: Hallucination resistance, knowledge boundary awareness

Each response is auto-tagged with behavioral metadata and compiled into a visual HTML report.

## Setup

```bash
# Install bdistill
pip install bdistill  # or: pip install -e ".[mcp]" from source

# Add MCP server to your tool
claude mcp add bdistill bdistill-mcp  # Claude Code
# Or copy .mcp.json to your project for Cursor/VS Code
```

## Usage

In Claude Code:
```
/xray                          # Full behavioral probe (30 questions)
/xray --dimensions refusal     # Probe just one dimension
/xray-report                   # Generate report from completed probe
```

In any tool with MCP:
```
"Run a behavioral self-probe"
"X-ray your own behavior"
"Test your refusal boundaries"
"Generate a behavioral report"
```

## Output

A styled HTML report showing:
- Refusal rate, hedge rate, chain-of-thought usage
- Per-dimension breakdown with bar charts
- Notable response examples with behavioral tags
- Actionable insights ("you already show CoT 85% of the time, no need to prompt for it")

登录后下载/点赞/收藏 ❤ 5 | ★ 0

behavioral-xray

目录结构

SKILL.md

举报内容

提示