Token导航 LogoToken导航

Agent Skill · 研究检索

knowledge(知识)

knowledge 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。

knowledge 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 24,147。

总安装 4,088周安装 167GitHub Stars 321下载 1,323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:knowledge(知识)
来源仓库:https://github.com/boshu2/agentops
仓库路径:skills/knowledge
安装命令:
npx skills add https://github.com/boshu2/agentops --skill knowledge
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/boshu2/agentops --skill knowledge

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。

knowledge 用于检索历史工作成果,支持关键词搜索与文件提取。适合在 Codex、Claude、Cursor、Gemini CLI 中查找过往解决方案。优先使用 ao search 命令,其次扫描 .agents/learnings/ 等本地目录。结果以文件形式返回,需人工判断适用性而非直接采纳结论。

knowledge 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/knowledge

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/boshu2/agentops --skill knowledge;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

knowledge 怎么安装?

在终端执行 npx skills add https://github.com/boshu2/agentops --skill knowledge。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

knowledge 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。

knowledge 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。

knowledge 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。

knowledge 用的人多吗?

公开数据是累计安装 4,088 次,近一周安装 167 次,下载 1,323 次,GitHub 321 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。

knowledge 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
knowledge

知识 · 本页

4,088321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

Knowledge Skill

YOU MUST EXECUTE THIS WORKFLOW. Do not just describe it.

Find and retrieve knowledge from past work.

Execution Steps

Given /knowledge <query>:

Step 1: Search with ao CLI (if available)

ao search "<query>" --limit 10 2>/dev/null

If results found, read the relevant files.

Step 2: Search.agents/ Directory

# Search learnings
grep -r "<query>" .agents/learnings/ 2>/dev/null | head -10

# Search patterns
grep -r "<query>" .agents/patterns/ 2>/dev/null | head -10

# Search research
grep -r "<query>" .agents/research/ 2>/dev/null | head -10

# Search retros
grep -r "<query>" .agents/retros/ 2>/dev/null | head -10

Step 3: Search Plans

# Local plans
grep -r "<query>" .agents/plans/ 2>/dev/null | head -10

# Global plans
grep -r "<query>" ~/.claude/plans/ 2>/dev/null | head -10

Step 3.5: Search Global Patterns

# Global patterns (cross-repo knowledge)
grep -r "<query>" ~/.claude/patterns/ 2>/dev/null | head -10

Global patterns contain knowledge promoted from any repository via /learn --global. These are high-confidence, cross-project learnings.

Step 3.6: Search Global Learnings

# Global learnings (cross-repo abstracted knowledge)
grep -r "<query>" ~/.agents/learnings/ 2>/dev/null | head -10

Global learnings are abstracted, transferable insights promoted from repo-specific learnings via /learn --promote or classified as cross-cutting by /retro.

Step 3.7: Search Global Patterns (new location)

# Global patterns (new location, cross-repo)
grep -r "<query>" ~/.agents/patterns/ 2>/dev/null | head -10

Step 4: Use Semantic Search (if MCP available)

Tool: mcp__smart-connections-work__lookup
Parameters:
  query: "<query>"
  limit: 10

Step 5: Read Relevant Files

For each match found, use the Read tool to get full content.

Step 6: Synthesize Results

Combine findings into a coherent response:

  • What do we know about this topic?
  • What learnings are relevant?
  • What patterns apply?
  • What past decisions were made?

Step 7: Report to User

Present the knowledge found:

  1. Summary of findings
  2. Key learnings (with IDs)
  3. Relevant patterns
  4. Links to source files
  5. Confidence level (how much we know)

Knowledge Locations

TypeLocationFormat
Learnings.agents/learnings/Markdown
Patterns.agents/patterns/Markdown
Research.agents/research/Markdown
Retros.agents/retros/Markdown
Plans.agents/plans/Markdown
Global Plans~/.claude/plans/Markdown
Global Learnings~/.agents/learnings/Cross-repo abstracted learnings
Global Patterns~/.agents/patterns/Cross-repo reusable patterns
Legacy Patterns~/.claude/patterns/Read-only fallback (deprecated for writes)

Key Rules

  • Search multiple locations - knowledge may be scattered
  • Use ao CLI first - semantic search is better
  • Fall back to grep - if ao not available
  • Read full files - don't just report matches
  • Synthesize - combine findings into useful answer

Example Queries

/knowledge authentication    # Find auth-related learnings
/knowledge "rate limiting"   # Find rate limit patterns
/knowledge kubernetes        # Find K8s knowledge
/knowledge "what do we know about caching"

Examples

Finding Past Learnings

User says: /knowledge "error handling patterns"

What happens:

  1. Agent tries ao search "error handling patterns", finds 3 matches
  2. Agent searches .agents/learnings/ with grep, finds 5 additional matches
  3. Agent searches .agents/patterns/ for related patterns, finds 2 matches
  4. Agent reads all matched files using Read tool
  5. Agent synthesizes findings into coherent response
  6. Agent reports: "We have 5 learnings about error handling: L1 (always wrap errors), L3 (use typed errors), L12 (log before returning), L15 (context propagation), L22 (retry with backoff)"
  7. Agent provides links to source files and confidence level: high (multiple confirmations)

Result: Complete knowledge synthesis with 5 specific learnings and 2 related patterns, all with source citations.

Querying Without ao CLI

User says: /knowledge "database migrations"

What happens:

  1. Agent tries ao search, command not found
  2. Agent falls back to grep search across .agents/ directories
  3. Agent finds 2 matches in learnings, 1 in research, 0 in patterns
  4. Agent reads matched files
  5. Agent synthesizes: "Limited knowledge found. L8 recommends using transaction-wrapped migrations. Research doc from 2026-01-20 analyzed migration tools."
  6. Agent reports medium confidence (only 2 sources)

Result: Knowledge found despite missing ao CLI, with appropriate confidence level based on source count.

Troubleshooting

ProblemCauseSolution
No results foundQuery too specific or knowledge not yet capturedBroaden search terms. Try synonyms. Check if topic was covered in recent work but retro not yet run. Suggest running /retro to extract recent learnings.
Too many results (overwhelming)Very broad query termNarrow query with more specific terms. Filter by date: search only recent learnings. Use semantic search (ao CLI) for better ranking if available.
Results lack contextGrep matches found but files don't address queryRead full files, not just matching lines. Synthesize from surrounding context. May need to trace back to original research with /trace.
Confidence level unclearMixed or contradictory sourcesReport conflicting information explicitly. Note which sources agree/disagree. Suggest running /research to investigate further if critical.

平台分布

Codex

32.96%
按下载量换算436

Claude

32.82%
按下载量换算434

Cursor

18.17%
按下载量换算240

Gemini CLI

8.61%
按下载量换算114

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