Skip to content
Back to skills

Grilling

ASecurity

适用于用户想对计划、设计、决策或想法进行高强度压力测试,或明确要求 grill、拷问、追问到底、深挖问题时。Relentlessly interview the user about a plan, design, decision, or idea in rounds, resolving the current decision frontier until both sides share a precise understanding. Use for stress-testing thinking, not routine clarification. 中文关键词:拷问、追问到底、深挖问题、压力测试、挑刺、决策树、grill me。

  • 3 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added September 29, 2026
ai-agentstesting

Security analysis

A100/100

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

Scanned September 29, 2026

npx -y skills add madnpc/skills --skill grilling --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Grilling?

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

Security grade badge for Grilling
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/madnpc-grilling/badge)](https://www.skillsdirectory.com/skills/madnpc-grilling)

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: grilling
description: "适用于用户想对计划、设计、决策或想法进行高强度压力测试,或明确要求 grill、拷问、追问到底、深挖问题时。Relentlessly interview the user about a plan, design, decision, or idea in rounds, resolving the current decision frontier until both sides share a precise understanding. Use for stress-testing thinking, not routine clarification. 中文关键词:拷问、追问到底、深挖问题、压力测试、挑刺、决策树、grill me。"
---

# Grilling

通过分轮访谈,把计划、设计、决策或想法绘制为一棵**决策树**:每个决策都会分叉出依赖它的后续决策。

## 输出语言

默认使用简体中文;用户明确指定其他语言时遵从其要求。代码、命令、产品名和必须保持原文的引用不强行翻译。

## 访谈循环

**决策前沿(frontier)**是所有前置决策已经确定、现在无需猜测就能回答的问题。按轮推进:每一轮询问整个当前前沿,为每个问题编号并给出建议答案,然后等待用户回答,再重新计算下一轮前沿。

每轮格式如下:

```markdown
❓ **Q1** - **<问题标题>**:<问题正文;可包含多个选项和必要说明>

➡️ <建议答案及简短理由>

---

❓ **Q2** - **<问题标题>**:<问题正文;可包含多个选项和必要说明>

➡️ <建议答案及简短理由>
```

遵守以下规则:

1. 从当前对话和可访问的环境中识别访谈对象、已有约束、未决事项及它们之间的依赖,形成决策树。
2. 同一轮只包含当前前沿:如果一个问题依赖本轮另一个尚未回答的问题,把它留到下一轮。
3. 每个问题都给出建议答案及简短理由;建议必须基于已知目标和约束,不要假装存在并不具备的确定性。
4. 能通过文件、代码或工具查明的事实,先自行查证,不要把事实查询推给用户。需要独立调查且环境支持子代理时,派出子代理查明;这项调查只阻塞依赖它的分支,其余前沿问题照常询问。
5. 每轮回答都会重塑决策树。对含糊回答、隐含假设、矛盾、边界情况和未说明的取舍继续追问,并据此重新计算前沿,不要机械执行预设问卷。

## 完成条件

只有当前沿为空、重要分支全部访问、依赖与约束均已明确且没有静默假设时,才总结双方的共同理解并请用户确认。

在用户明确确认已达成共识之前,不要实施方案、写代码、修改文件或执行讨论中的决定。

Files in this skill

  • SKILL.md2.5 KB
  • agents/openai.yaml291 B

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.