Skip to content
Agentmart
Submit a tool

auto-harness-santa

执行 Santa Method 双独立对抗验证。Use when reviewing high-risk changes, production deployments, or complex logic before shipping.

Skill by duoglas

IndexedMIT★ 39
Write a review

Chalk Labs

Need help deploying auto-harness-santa?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/duoglas-auto-harness-santa

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

执行 Santa Method 双独立对抗验证。Use when reviewing high-risk changes, production deployments, or complex logic before shipping.

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/auto-harness-santa

What's inside

Skill name
auto-harness-santa
Description
执行 Santa Method 双独立对抗验证。Use when reviewing high-risk changes, production deployments, or complex logic before shipping.
License
MIT
Instructions
126 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Harness Santa

双独立 Reviewer 对抗验证——两个 Agent 独立审查,都通过才放行。

## 何时使用

- 高风险代码变更(安全、支付、核心逻辑)
- 生产部署前
- 复杂算法或多组件集成
- 用户说"Santa review"或"对抗验证"

## 何时不用

- 低风险改动(文档、配置、格式化)
- 有充分自动化测试覆盖的简单变更
- 探索性原型

## 执行流程

### Phase 1: Make a List(准备)

收集待审查的材料:
- 代码变更(git diff 或文件内容)
- 需求规格(如果有)
- 审查 Rubric(见下方模板)

### Phase 2: Check It Twice(双独立审查)

并行启动两个独立 Reviewer Agent:

**Reviewer A(Claude Agent):**
```
你是独立的质量审查者。你没有看过其他任何审查意见。

## 待审查内容
{代码变更}

## 审查标准
{Rubric}

## 指令
逐项对照标准检查。对每一项:
- PASS: 完全满足,无问题
- FAIL: 发现具体问题(引用具体位置)

输出 JSON:
{
  "verdict": "PASS" | "FAIL",
  "checks": [{"criterion": "...", "result": "PASS|FAIL", "detail": "..."}],
  "critical_issues": ["..."],
  "suggestions": ["..."]
}

严格检查。你的工作是发现问题。
```

**Reviewer B(可选用 Codex `/codex:adversarial-review`,或另一个 Claude Agent):**
同样的 Rubric,同样的代码变更,独立上下文。

**关键不变量:**
- 两个 Reviewer 互不可见(上下文隔离)
- 使用相同的 Rubric
- 并行执行

### Phase 3: Naughty or Nice(判决)

```
Reviewer A: PASS  AND  Reviewer B: PASS  →  NICE  →  放行
其他情况                                →  NAUGHTY  →  Phase 4
```

一个发现问题就算 NAUGHTY。

### Phase 4: Fix Until Nice(修复循环)

```
收集两个 Reviewer 的所有 issues
    ↓
派 Fix Agent 修复(只修 flagged issues,不做额外重构)
    ↓
重新启动两个全新 Reviewer Agent 审查修复后的代码
    ↓
最多 3 轮。超过 3 轮 → 升级给人工处理。
```

**关键:** 每轮修复后用全新 Agent 重审,防止锚定偏差。

## Rubric 模板

| 检查项 | PASS 条件 | FAIL 信号 |
|--------|----------|----------|
| 功能正确性 | 所有 spec 需求被覆盖 | 遗漏需求项 |
| 无幻觉 | 无虚构的 API/函数/URL | 引用不存在的东西 |
| 安全性 | 无硬编码密钥,输入有验证 | 发现敏感信息 |
| 一致性 | 无自相矛盾 | A 处说 X,B 处说 non-X |
| 技术正确性 | 代码可编译运行,逻辑正确 | 语法错误、逻辑 bug |
| 错误处理 | 异常路径有处理 | 缺少 error handling |

根据项目补充领域特定的检查项。

## 指标

- **First-pass rate**: 第一轮通过率(目标 >70%)
- **Mean iterations**: 平均修复轮数(目标 <1.5)
- **Reviewer agreement**: 两个 Reviewer 同时发现的问题占比
- **Escape rate**: Santa 通过但上线后发现的问题(目标 0)

## 跨模型扩展

Reviewer A 用 Claude,Reviewer B 用 Codex(`/codex:adversarial-review`)。不同模型有不同盲区,交叉审查进一步降低逃逸率。

## AI 工具内测试准出协议

只要任务涉及代码变更,AI 不能等用户提醒才验证。按下面顺序做:

1. 先判断风险等级:low / medium / high / release。
E2E PASS 不等于充分;如果只是 echo ok、空脚本、只 smoke、或没覆盖本次风险,用户报告要先说“现在还不能交付”,再说明测到了什么、没测到什么、下一步补什么;机器状态放最后,例如:机器状态:NOT_SUFFICIENT。DEGRADED 不能说成 PASS。
2. 识别测试能力:单测、lint、coverage、E2E、runtime smoke。
3. medium / high / release 任务必须有 E2E 证据;只有 low 小改可以不强制 E2E;找不到 E2E 入口时,先生成计划或只问一个具体启动问题。

From duoglas/simple-harness-kit, MIT.

More from duoglas/simple-harness-kit

Signals

Stars
39
Forks
3
Open issues
0
Last commit (this folder)
Jun 5, 2026 (5 months ago)
Last release
Oct 9, 2026 (26 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from duoglas

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get auto-harness-santa running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying auto-harness-santa?

CallGet help
Cite this page

auto-harness-santa skill. Agent Mart. https://agentmart.dev/skills/duoglas-auto-harness-santa (last updated 2026-10-24)