Token导航 LogoToken导航

Agent Skill · 研究检索

dissent(dissent 搜索)

dissent 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

dissent 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 35,826。

总安装 2,022周安装 81GitHub Stars 1下载 654核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:dissent(dissent 搜索)
来源仓库:https://github.com/open-horizon-labs/skills
仓库路径:skills/dissent
安装命令:
npx skills add https://github.com/open-horizon-labs/skills --skill dissent
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/open-horizon-labs/skills --skill dissent

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

结构化异议机制,用于强化关键决策的质量保障。在架构锁定、API 发布等重大节点前主动寻求反证。采用钢铁人原则重构对方论点,非攻击性质疑。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。适用于高 stakes 且团队收敛过快时的决策检验场景。

dissent 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/dissent

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/open-horizon-labs/skills --skill dissent;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

dissent 怎么安装?

在终端执行 npx skills add https://github.com/open-horizon-labs/skills --skill dissent。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

dissent 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

dissent 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

dissent 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

dissent 用的人多吗?

公开数据是累计安装 2,022 次,近一周安装 81 次,下载 654 次,GitHub 1 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

dissent 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
dissent

dissent 搜索 · 本页

2,0221需要联网未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

/dissent

Structured disagreement that strengthens decisions. Find flaws before the one-way door closes.

Dissent is not attack. It's actively seeking reasons you might be wrong.

When to Use

Invoke /dissent when: about to lock in a one-way door (architecture, public API, major hires), confidence is high but stakes are higher, team is converging too quickly, you're defending a position, or the path forward seems obvious.

Do not use when: Gathering initial options or brainstorming. Dissent stress-tests decisions, not generates them.

The Dissent Process

Step 1: Steel-Man the Current Approach

Before attacking, articulate the position:

"The current approach is [approach]. The reasoning is [reasoning]. The expected outcome is [outcome]. This is the best version of this position."

If you can't state it charitably, you don't understand it well enough to challenge it.

Step 2: Seek Contrary Evidence

  • What data would prove this approach wrong?
  • Who disagrees? What's their argument?
  • What similar approaches have failed elsewhere? Why?
  • What are we ignoring because it's inconvenient?
"If this approach were wrong, what would we expect to see? Are we seeing any of that?"

Step 3: Pre-Mortem

Imagine it's six months from now and this decision failed. Work backward:

"This failed because [reason]. The warning signs we ignored were [signs]. The assumption that broke was [assumption]."

Generate at least three plausible failure scenarios:

  1. Technical failure — it doesn't work as expected
  2. Adoption failure — it works but nobody uses it / changes nothing
  3. Opportunity cost — it works but we should have done something else

Step 4: Surface Hidden Assumptions

Every decision rests on unstated assumptions. Find them:

  • What are we assuming about the user/customer?
  • What are we assuming about the system/codebase?
  • What are we assuming about timeline/resources?
  • What are we assuming won't change?

For each:

Assumption: [what we're taking for granted]
Evidence: [what supports this]
Risk if wrong: [what happens]
Test: [how to validate before committing]

Step 5: Decide

  • PROCEED — no critical flaws found; decision strengthened
  • ADJUST — issues surfaced that can be addressed; modify the approach
  • RECONSIDER — fundamental problems revealed; back to solution space
"PROCEED: The counter-argument is [X], but it's addressed by [Y]. The key assumption is [Z], which we've validated by [how]."

ADR Generation

If the decision is a one-way door and you recommend PROCEED or ADJUST, offer to create an Architecture Decision Record. The dissent report maps directly:

Dissent SectionADR Section
Decision under reviewTitle
Steel-Man PositionContext
Contrary Evidence + Pre-MortemOptions Considered
Hidden AssumptionsConsequences
Decision + ReasoningDecision

Write to docs/adr/NNNN-<decision-slug>.md if accepted.

Output Format

## Dissent Report

**Decision under review:** [what's being challenged]
**Stakes:** [why this matters]
**Confidence before dissent:** [HIGH/MEDIUM/LOW]

### Steel-Man Position
[The best version of the current approach]

### Contrary Evidence
1. [Evidence point 1]
2. [Evidence point 2]
3. [Evidence point 3]

### Pre-Mortem Scenarios
1. **[Failure mode]:** [How this could fail]
2. **[Failure mode]:** [How this could fail]
3. **[Failure mode]:** [How this could fail]

### Hidden Assumptions
| Assumption | Evidence | Risk if Wrong | Test |
|------------|----------|---------------|------|
| [Assumption 1] | [Evidence] | [Risk] | [Test] |

### Decision

**Recommendation:** [PROCEED / ADJUST / RECONSIDER]
**Reasoning:** [Why]
**If ADJUST:** [Specific modifications]
**Confidence after dissent:** [HIGH/MEDIUM/LOW]
**Create ADR?** [Yes - offer path / No - not a one-way door]

Session Persistence

Reads: Aim, Problem Statement, Solution Space — context for the decision being challenged.

Writes: Dissent report:

## Dissent
**Updated:** <timestamp>
**Decision:** [PROCEED | ADJUST | RECONSIDER]

[dissent report content]

Position in Framework

Combines with: /solution-space (challenge the recommendation), /problem-statement (challenge the framing), /execute (before one-way doors). Leads to: PROCEED (continue with confidence), ADJUST (modify approach), or RECONSIDER (back to solution-space). This is not a phase: Dissent is an overlay you invoke when stakes are high.

平台分布

Codex

36.3%
按下载量换算237

Claude

31.32%
按下载量换算205

Cursor

19.89%
按下载量换算130

Gemini CLI

9.22%
按下载量换算60

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills