Token导航 LogoToken导航

Agent Skill · 研究检索

refine(refine 搜索)

refine 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。

refine 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,权限为只读;许可证为 MIT,暂无第三方扫描结果,按安装量在研究检索类的 73,675 个 Skill 里排第 47,060。

总安装 751周安装 31GitHub Stars 公开资料未说明下载 246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
MIT

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:refine(refine 搜索)
来源仓库:https://github.com/duc01226/easyplatform
仓库路径:skills/refine
安装命令:
npx skills add duc01226/easyplatform --skill "refine"
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。

AgentSkills.tonpx skills
npx skills add duc01226/easyplatform --skill "refine"

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。

refine 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。它支持通过自然语言描述目标信息,自动匹配并返回结构化数据,适用于研究、开发或日常信息整理场景。

安装后可通过 npx skills add duc01226/easyplatform --skill "refine" 命令集成到宿主环境,具体调用方式请参考原始 README。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操作。

refine 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/refine

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 duc01226/easyplatform --skill "refine";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

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

refine 怎么安装?

在终端执行 npx skills add duc01226/easyplatform --skill "refine"。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

refine 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。

refine 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

refine 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

refine 用的人多吗?

公开数据是累计安装 751 次,近一周安装 31 次,下载 246 次。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。

refine 安全吗?

没有查到第三方扫描结果,安全性无法从公开资料判断,安装前请自行审查源码、依赖和命令行为。

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。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 没有第三方扫描结果,安全性只能靠自己读源码判断。
  • 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。
  • 只想先试试效果、不希望它改动本地文件的人。

不太适合

  • 要求有安全扫描结论才肯安装的团队。
  • 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
refine

refine 搜索 · 本页

751—只读MIT未收录

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

Idea Refinement to PBI

Transform captured ideas into actionable Product Backlog Items using Business Analysis best practices, Hypothesis-Driven Development, and domain research.

When to Use

  • Idea artifact ready for refinement
  • Need to validate problem hypothesis before building
  • Converting concept to implementable item
  • Adding acceptance criteria to requirements
  • Researching domain/market context for new features

Quick Reference

Workflow Overview

PhaseNameKey ActivityOutput
1Idea IntakeLoad artifact, detect moduleContext loaded
2Domain ResearchWebSearch market/competitorsResearch summary
3Problem HypothesisValidate problem existsConfirmed hypothesis
4ElicitationApply BABOK techniquesRequirements extracted
5Acceptance CriteriaWrite BDD scenariosGIVEN/WHEN/THEN
6PrioritizationApply RICE/MoSCoWPriority assigned
7ValidationInterview userAssumptions confirmed
8PBI GenerationCreate artifactPBI file saved

Output

  • Path: team-artifacts/pbis/{YYMMDD}-pbi-{slug}.md
  • ID Pattern: PBI-{YYMMDD}-{NNN}

Phase 1: Idea Intake & Context Loading

Process

  1. Read idea artifact from path or find by ID in team-artifacts/ideas/
  2. Extract: problem statement, value proposition, target users, scope
  3. Check module field in frontmatter for BravoSUITE domain

Module Detection

If module present: Load business feature context (see BravoSUITE Integration)

If module absent:

  1. Run Glob("docs/business-features/*/README.md") to discover modules
  2. Analyze idea text for module keywords
  3. Prompt if ambiguous: "Which BravoSUITE module?" + list + "None"

Skip conditions: Infrastructure ideas, cross-cutting concerns


Phase 2: Domain Research

When to Trigger

  • New market/domain unfamiliar to team
  • Competitive landscape unclear
  • Industry best practices needed
  • User explicitly requests --research

Skip When

  • Internal tooling with no market equivalent
  • Well-understood domain
  • Time-constrained refinement

Process

  1. Extract key domain terms from idea
  2. Use WebSearch for context:
Query TypeTemplate
Market trends"{domain} market trends 2026"
Competitors"{domain} software solutions comparison"
Best practices"{feature-type} best practices UX"
Similar solutions"how {competitor} handles {feature}"
  1. Summarize findings (max 3 bullets)

Output Template

## Domain Research Summary
- **Market context:** {1-sentence finding}
- **Competitor landscape:** {key players, gaps identified}
- **Best practices:** {relevant pattern to adopt}
- **Sources:** {links}

Phase 3: Problem Hypothesis Validation

Why This Matters

42% of startups fail due to no market need (CB Insights). Validate before building.

Problem Hypothesis Template

**We believe** [target users/persona]
**Experience** [specific problem]
**Because** [root cause]
**We'll know this is true when** [validation metric/evidence]

Value Hypothesis Template

**We believe** [feature/solution]
**Will deliver** [value/benefit]
**To** [target users]
**We'll know we're right when** [success metric]

Validation Process

  1. Draft hypothesis from idea content
  2. Use AskUserQuestion to validate:

- "Is this the core problem we're solving?" - "Who exactly experiences this? How often?" - "What evidence do we have this problem exists?"

  1. If validated, proceed to Phase 4
  2. If invalidated, return idea for clarification

Skip When

  • --skip-hypothesis flag provided
  • Idea already contains validated hypothesis
  • Bug fix or technical debt (no new problem)

Phase 4: Requirements Elicitation (BABOK Core 5)

Select technique based on context:

1. Interviews

When: Deep individual insights needed, stakeholder perspectives vary

Process:

  1. Prepare open-ended questions (why, how, what-if)
  2. Active listening - don't lead the witness
  3. Follow-up on unexpected answers
  4. Document verbatim quotes for evidence

Output: Stakeholder needs, pain points, constraints

2. Workshops

When: Group consensus needed, complex requirements, multiple stakeholders

Process:

  1. Define clear agenda and timebox (90 min max)
  2. Use facilitator (neutral party)
  3. Capture all voices (round-robin, silent voting)
  4. Document decisions and dissent

Output: Prioritized requirements, consensus decisions

3. Document Analysis

When: Existing systems/processes to understand, regulatory requirements

Process:

  1. Gather artifacts: specs, manuals, reports, existing code
  2. Extract implicit requirements
  3. Note gaps and inconsistencies
  4. Cross-reference with stakeholder input

Output: As-is state, compliance requirements, gaps

4. Observation (Job Shadowing)

When: Understand real workflow, users can't articulate needs

Process:

  1. Shadow users in their environment
  2. Note workarounds and pain points
  3. Don't interrupt or suggest - just observe
  4. Ask clarifying questions afterward

Output: Actual vs stated workflow, hidden requirements

5. Prototyping

When: Visual validation needed, UI/UX requirements unclear

Process:

  1. Start low-fidelity (sketches, wireframes)
  2. Iterate based on feedback
  3. Increase fidelity as requirements stabilize
  4. Document design decisions

Output: Validated UI requirements, interaction patterns


Phase 5: Acceptance Criteria (BDD Format)

Standard Format

Scenario: {Descriptive title}
  Given {precondition/context}
    And {additional context}
  When {action/trigger}
    And {additional action}
  Then {expected outcome}
    And {additional verification}

Best Practices

PracticeDescription
Single trigger"When" clause has ONE action
3 scenarios minimumHappy path, edge case, error case
No implementation detailsFocus on behavior, not how
Testable outcomes"Then" must be verifiable
Stakeholder languageNo technical jargon

Example Scenarios

Scenario: Employee creates goal with valid data
  Given employee has permission to create goals
    And employee is on the goal creation page
  When employee submits goal form with all required fields
  Then goal is created with status "Draft"
    And goal appears in employee's goal list

Scenario: Goal creation fails with missing required field
  Given employee is on the goal creation page
  When employee submits form without title
  Then validation error "Title is required" is displayed
    And goal is not created

Scenario: Manager reviews subordinate goal
  Given manager has direct reports
    And subordinate has submitted goal for review
  When manager opens goal review page
  Then subordinate's goal is visible with "Pending Review" status

BravoSUITE Test Case Format

For domain features, use:

  • Format: TC-{MOD}-{FEATURE}-XXX (e.g., TC-GRO-GOAL-001)
  • Evidence: file:line format
  • See business-analyst skill for detailed patterns

Phase 6: Prioritization & Estimation

Quick RICE Score

Score = (Reach × Impact × Confidence) / Effort

Reach: Users affected per quarter (100, 500, 1000+)
Impact: 0.25 (minimal) | 0.5 (low) | 1 (medium) | 2 (high) | 3 (massive)
Confidence: 0.5 (low) | 0.8 (medium) | 1.0 (high)
Effort: Person-days (1, 3, 5, 10, 20)

MoSCoW Categories

CategoryMeaningAction
Must HaveCritical, non-negotiableInclude in MVP
Should HaveImportant but not vitalPlan for release
Could HaveNice to have, low effortIf time permits
Won't HaveOut of scope this cycleDocument for future

Effort Estimation

T-ShirtDaysWhen to Use
XS0.5-1Config change, simple fix
S1-2Single component, clear scope
M3-5Multiple components, some unknowns
L5-10Cross-cutting, integration needed
XL10+Epic - break down further

Phase 7: Validation Interview (MANDATORY)

After drafting PBI, validate with user.

Question Categories

CategoryExample Question
Assumptions"We assume X is true. Correct?"
Scope"Should Y be included or explicitly excluded?"
Dependencies"This requires Z. Is that available?"
Edge Cases"What happens when data is empty/null?"
Business Impact"Will this affect existing reports/workflows?"
Entities"Create new entity or extend existing X?"

Process

  1. Generate 3-5 questions from assumptions, scope, dependencies
  2. Use AskUserQuestion tool to interview
  3. Document in PBI under ## Validation Summary
  4. Update PBI based on answers

Validation Output Format

## Validation Summary

**Validated:** {date}

### Confirmed Decisions
- {decision}: {user choice}

### Assumptions Confirmed
- {assumption}: Confirmed/Modified

### Open Items
- [ ] {follow-up items}

This step is NOT optional - always validate before marking complete.


Phase 8: PBI Artifact Generation

Save Location

team-artifacts/pbis/{YYMMDD}-pbi-{slug}.md

Required Sections

  1. Frontmatter (ID, source_idea, priority, effort, status, module)
  2. Description
  3. Business Value
  4. Problem Hypothesis (from Phase 3)
  5. Related Business Rules (if domain)
  6. Acceptance Criteria (from Phase 5)
  7. Out of Scope
  8. Dependencies
  9. Validation Summary (from Phase 7)

Template Reference

See: team-artifacts/templates/pbi-template.md


Definition of Ready Checklist

Before marking PBI as "Ready":

CriterionCheck
IndependentNo blocking dependencies on other PBIs
NegotiableDetails can still be refined with team
ValuableClear user/business value articulated
EstimableTeam can estimate effort (XS-XL)
SmallCan complete in single sprint
TestableHas 3+ GIVEN/WHEN/THEN scenarios
Problem ValidatedHypothesis confirmed in Phase 3
Domain ContextBR/entity context loaded (if BravoSUITE)
Stakeholder AlignedValidation interview completed

BravoSUITE Integration

For domain-related PBIs, load business feature context.

Module Detection

Reference: .claude/skills/shared/module-detection-keywords.md

Business Rules Extraction

  1. Read docs/business-features/{module}/README.md
  2. Identify related feature doc
  3. Extract BR-{MOD}-XXX rules from "Business Rules" section
  4. Note conflicts with new requirements

Entity Context

  1. Read .ai.md companion file
  2. Extract entity names, properties, relationships
  3. Identify key expressions for queries

Token Budget

Target 8-12K tokens for feature context:

  • Module README: ~2K tokens
  • Feature doc sections: 3-5K per feature

Detailed patterns: See business-analyst skill


Anti-Patterns to Avoid

Anti-PatternBetter Approach
Refining vague ideasReturn to /idea for clarification
Skipping hypothesis validationAlways run Phase 3 for new features
Solution-first thinkingStart with problem, not solution
Generic acceptance criteriaUse GIVEN/WHEN/THEN with specifics
Ignoring domain contextLoad BravoSUITE docs if applicable
Too large PBI (XL+)Break into smaller items
Missing "Out of Scope"Explicitly list exclusions
Assuming instead of askingRun validation interview

Templates Quick Reference

Problem Hypothesis

**We believe** {users} **experience** {problem} **because** {cause}.
**Validation:** {metric/evidence}

Value Hypothesis

**We believe** {solution} **will deliver** {value} **to** {users}.
**Success metric:** {how we measure}

Acceptance Criteria

Scenario: {Title}
  Given {context}
  When {action}
  Then {outcome}

Related

  • Role Skill: business-analyst (detailed patterns)
  • Command: /refine
  • Input: /idea output
  • Next Step: /story, /test-spec, /design-spec
  • Prioritization: /prioritize

Triggers

Activates on: refine, refinement, pbi, backlog item, acceptance criteria, hypothesis, validate idea


Task Management Protocol: - Always plan and break work into many small todo tasks - Always add a final review todo task to verify work quality and identify fixes/enhancements

平台分布

Claude Code

27.44%
按下载量换算68

windsurf

24.12%
按下载量换算59

OpenCode

17.89%
按下载量换算44

Codex

12.5%
按下载量换算31

Antigravity

7.98%
按下载量换算20

Gemini CLI

3.7%
按下载量换算9

安全审计

暂无安全审计结果可展示。

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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