Token导航 LogoToken导航

Agent Skill · 研究检索

opus-extended-thinking(作品延伸思维)

opus-extended-thinking 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。

opus-extended-thinking 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,权限为调用外部服务;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 70,604。

总安装 210周安装 9GitHub Stars 3下载 73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:opus-extended-thinking(作品延伸思维)
来源仓库:https://github.com/frankxai/arcanea
仓库路径:skills/opus-extended-thinking
安装命令:
npx skills add https://github.com/frankxai/arcanea --skill opus-extended-thinking
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/frankxai/arcanea --skill opus-extended-thinking

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。

扩展 Opus 模型的推理深度与上下文记忆能力。适用于长文档摘要、复杂逻辑链推导等高阶认知任务。通过分块处理与中间状态缓存提升超长输入处理能力。消耗更多 token 且响应时间延长,建议按需启用此模式。opus-extended-thinking 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/opus-extended-thinking

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/frankxai/arcanea --skill opus-extended-thinkin…;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

    它需要调用外部服务,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

opus-extended-thinking 怎么安装?

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

opus-extended-thinking 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。

opus-extended-thinking 需要什么权限?

权限级别是调用外部服务。该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

opus-extended-thinking 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。

opus-extended-thinking 用的人多吗?

公开数据是累计安装 210 次,近一周安装 9 次,下载 73 次,GitHub 3 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。

opus-extended-thinking 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
opus-extended-thinking

作品延伸思维 · 本页

2103调用外部服务未明确通过

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

Claude Opus 4.5 Extended Thinking

This skill guides optimal use of Claude Opus 4.5's extended thinking capabilities - the model's ability to reason deeply before responding, leading to more thoughtful, accurate, and nuanced outputs.


Understanding Extended Thinking

What It Is

Extended thinking allows Opus 4.5 to:

  • Deliberate longer before producing output
  • Consider multiple perspectives and approaches
  • Self-critique and refine reasoning
  • Handle complex multi-step problems
  • Produce more thoughtful, less reactive responses

When It Activates

Extended thinking engages automatically when:

  • Problems require multi-step reasoning
  • Questions have nuance or ambiguity
  • Tasks require weighing tradeoffs
  • Complex analysis or synthesis is needed
  • Creative work benefits from deliberation

Prompt Patterns for Deep Thinking

Pattern 1: Explicit Deliberation Request

Before answering, take time to:
1. Consider the problem from multiple angles
2. Identify potential pitfalls or edge cases
3. Weigh different approaches
4. Then provide your best response

Question: [Your complex question]

Pattern 2: Multi-Perspective Analysis

I need a thorough analysis of [topic].

Please consider:
- The technical perspective
- The business/practical perspective
- The user/human perspective
- Potential risks and downsides
- Long-term implications

Take your time to reason through each before synthesizing.

Pattern 3: Step-by-Step Reasoning

Walk me through your reasoning on this problem step by step.
Don't jump to conclusions - show your work.

Problem: [Complex problem]

For each step, explain:
- What you're considering
- Why you're making that choice
- What alternatives you rejected and why

Pattern 4: Devil's Advocate

I'm planning to [action/decision].

Before you help me execute this:
1. Steelman the opposing view - what's the best argument against?
2. What am I potentially missing or underweighting?
3. What would make this fail?
4. Only then, if it still seems right, help me proceed.

Pattern 5: Synthesis Request

I have [number] different sources/perspectives on [topic]:
[List them]

Please synthesize these into a coherent understanding that:
- Identifies where they agree
- Explains where they conflict and why
- Weighs the evidence/arguments
- Provides your reasoned conclusion

Task Types That Benefit Most

1. Strategic Planning

Prompt pattern:
"Help me think through [strategic decision].
Consider: current state, desired outcome, constraints,
multiple paths, risks of each, second-order effects.
Don't rush to a recommendation - reason through thoroughly first."

2. Code Architecture Decisions

Prompt pattern:
"I'm deciding between [Option A] and [Option B] for [system].
Before recommending, analyze:
- Scalability implications
- Maintenance burden
- Team capability fit
- Migration complexity
- What we might regret in 2 years"

3. Creative Development

Prompt pattern:
"I want to create [creative work] about [theme].
Before writing, explore:
- Different narrative approaches
- Tonal variations
- Structural options
- What makes this theme resonate
Then develop the direction that feels most promising."

4. Problem Diagnosis

Prompt pattern:
"[System/situation] is exhibiting [problem].
Don't jump to the obvious cause.
Consider:
- What are all possible causes?
- What evidence supports/refutes each?
- What's the most likely root cause?
- What would I need to verify to be sure?"

5. Research Synthesis

Prompt pattern:
"Help me understand [complex topic].
I need you to:
- Break down the key concepts
- Explain how they relate
- Identify what's well-established vs debated
- Note what's commonly misunderstood
- Synthesize into a coherent mental model"

Anti-Patterns to Avoid

❌ Rushing to Action

Bad: "Write me a function that does X"
Better: "I need a function that does X. Before coding,
        let's consider: edge cases, error handling,
        interface design, testing approach."

❌ Accepting First Answer

Bad: "What's the best database for my app?"
Better: "What database should I use for [specific requirements]?
        Consider multiple options, tradeoffs between them,
        and what factors should weight most heavily."

❌ Binary Thinking

Bad: "Should I do A or B?"
Better: "I'm choosing between A and B, but I'm open to
        alternatives I haven't considered. What's the
        best path forward given [context]?"

❌ Ignoring Uncertainty

Bad: "Tell me the answer to [complex question]"
Better: "Help me understand [complex question].
        Be explicit about what's certain vs uncertain,
        and what evidence supports different views."

Extended Thinking Triggers

Use these phrases to encourage deeper deliberation:

Analytical Triggers:

  • "Think through this carefully..."
  • "Consider all angles before responding..."
  • "What are the non-obvious implications?"
  • "What am I likely missing?"
  • "Reason through step by step..."

Critical Triggers:

  • "What could go wrong with this approach?"
  • "Steelman the opposing view..."
  • "What's the strongest argument against?"
  • "What assumptions am I making?"
  • "Where might this reasoning break down?"

Synthesis Triggers:

  • "How do these pieces fit together?"
  • "What's the underlying pattern?"
  • "Synthesize into a coherent view..."
  • "What's the essence of this?"
  • "How would you explain this to an expert?"

Output Quality Indicators

Signs extended thinking is working well:

  • Response acknowledges complexity and tradeoffs
  • Multiple perspectives are genuinely considered
  • Reasoning is visible, not just conclusions
  • Uncertainty is explicitly addressed
  • Edge cases and risks are surfaced proactively
  • Recommendations are nuanced, not absolute

Signs it might need more prompting:

  • Quick, surface-level response
  • Only one perspective considered
  • Confident without showing reasoning
  • Missing obvious complexities
  • Generic advice without context fit

Combining with Other Patterns

Extended Thinking + Agentic Execution

"First, think through the best approach to [task].
Consider alternatives and tradeoffs.
Once you have a clear plan, execute it autonomously,
but note any decision points where you chose between options."

Extended Thinking + Iterative Refinement

"Draft an initial approach to [problem].
Then critique your own draft - what could be better?
Revise based on your critique.
Show me both the final version and key improvements you made."

Extended Thinking + User Collaboration

"Analyze [situation] and identify the key decision points.
For each decision, explain the tradeoffs.
Before proceeding, I'll tell you my priorities,
then you can recommend the path that fits best."

When NOT to Use Extended Thinking

Some tasks benefit from quick, direct responses:

  • Simple factual questions
  • Straightforward code snippets
  • Clear instructions with no ambiguity
  • Tasks where speed matters more than depth
  • Follow-up execution after thinking is done

For these, don't add complexity. Just ask directly.


Model-Specific Notes (Opus 4.5)

Opus 4.5 specifically excels at:

  • Sustained reasoning chains - Can maintain coherence over long analytical sequences
  • Self-correction - Will catch and correct its own errors when given space
  • Nuance handling - Better at "it depends" answers with clear conditions
  • Multi-factor synthesis - Can hold many variables in consideration simultaneously
  • Creative-analytical blend - Can reason rigorously about creative work

Integration with FrankX System

Use extended thinking for:

  • Starlight Council decisions - Strategic synthesis across domains
  • Architecture planning - System design for Arcanea, FrankX projects
  • Content strategy - Weighing positioning, audience, platform fit
  • Creative development - Book outlines, course structures, music direction
  • Technical decisions - Framework choices, MCP design, agent orchestration

*Extended thinking is about quality of cognition, not length of output. The best responses come from genuine deliberation - use these patterns to unlock that capability.*

平台分布

Codex

37.01%
按下载量换算27

Claude

26.82%
按下载量换算20

Cursor

18.92%
按下载量换算14

Gemini CLI

9.57%
按下载量换算7

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

调用外部服务

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

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