Token导航 LogoToken导航TokenDH.com
研究检索需要联网github未标认证来源可访问许可证需确认审计未展示

ring%3apre-dev-researchRing%3apre 开发研究

Agent Skill

ring%3apre-dev-research 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。

总安装

881

周安装

36

GitHub Stars

180

下载量

282
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

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:ring%3apre-dev-research(Ring%3apre 开发研究)
来源仓库:https://github.com/lerianstudio/ring
仓库路径:skills/ring%3Apre-dev-research
安装命令:
npx skills add https://github.com/lerianstudio/ring --skill ring:pre-dev-research
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/lerianstudio/ring --skill ring:pre-dev-research

简介

ring%3apre-dev-research 用于查找、检索和筛选相关信息。

  • 它适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词或任务线索定位内容。
  • 可通过 npx skills add 命令从 GitHub 仓库安装,具体实现以原始 README 为准。
  • 安装前建议评估权限范围和潜在的网络访问或命令执行行为。
  • 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。

SKILL.md

Pre-Dev Research Skill (Gate 0)

Purpose: Gather comprehensive research BEFORE writing planning documents, ensuring PRDs and TRDs are grounded in codebase reality and industry best practices.

The Research-First Principle

Traditional:  Request → PRD → Discover problems during implementation
Research-First:  Request → Research → Informed PRD → Smoother implementation

Research prevents: Reinventing existing patterns, ignoring conventions, missing framework constraints, repeating solved problems

Step 1: Determine Research Mode

BLOCKING GATE: Before dispatching agents, determine the research mode.

ModeWhen to UseExample
greenfieldNo existing patterns"Add GraphQL API" (when project has none)
modificationExtending existing functionality"Add pagination to user list API"
integrationConnecting external systems"Integrate Stripe payments"

If unclear, ask:

Before starting research: Is this (1) Greenfield - new capability, (2) Modification - extends existing, or (3) Integration - connects external systems?

Mode affects agent priority:

  • Greenfield → Web research primary (best-practices, framework-docs)
  • Modification → Codebase research primary (repo-research)
  • Integration → All agents equally weighted

Step 2: Dispatch Research Agents

Run 4 agents in PARALLEL (single message, 4 Task calls):

AgentPrompt FocusMode
ring:repo-research-analystCodebase patterns for [feature]. Search docs/solutions/ knowledge base. Return file:line references.PRIMARY in modification
ring:best-practices-researcherExternal best practices for [feature]. Use Context7 + WebSearch. Return URLs.PRIMARY in greenfield
ring:framework-docs-researcherTech stack docs for [feature]. Detect versions from manifests. Use Context7. Return version constraints.PRIMARY in integration
ring:product-designerUser problem validation, personas, competitive UX analysis, design constraints for [feature]. Mode: ux-research.PRIMARY in greenfield, SECONDARY in others

Task invocation for product-designer:

Task(
  subagent_type="ring:product-designer",
  prompt="Analyze user needs for [feature]. Mode: ux-research. Return: problem validation, preliminary personas, competitive analysis, design constraints."
)

Step 2.5: Handle Topology Configuration

If TopologyConfig is provided (from command's topology discovery):

  1. Persist in research.md frontmatter:
---
feature: {feature-name}
gate: 0
date: {YYYY-MM-DD}
research_mode: greenfield | modification | integration
agents_dispatched: 4
topology:
  scope: fullstack | backend-only | frontend-only
  structure: single-repo | monorepo | multi-repo
  modules:  # Only if monorepo or multi-repo
    backend:
      path: {path}
      language: golang | typescript
    frontend:
      path: {path}
      framework: nextjs | react | vue
  doc_organization: unified | per-module
  api_pattern: direct | bff | other  # Only if scope=fullstack
---

Document Placement (based on topology.structure):

Structureresearch.md Location
single-repodocs/pre-dev/{feature-name}/research.md
monorepodocs/pre-dev/{feature-name}/research.md (root)
multi-repoWrite to BOTH: {backend.path}/docs/pre-dev/{feature-name}/research.md AND {frontend.path}/docs/pre-dev/{feature-name}/research.md

Multi-repo directory creation:

# Create directories in both repos
mkdir -p "{backend.path}/docs/pre-dev/{feature-name}"
mkdir -p "{frontend.path}/docs/pre-dev/{feature-name}"

Multi-repo sync note: Add this footer to research.md for multi-repo:

---
**Sync Status:** This document is maintained in both repositories.
- Primary: {backend.path}/docs/pre-dev/{feature-name}/research.md
- Mirror: {frontend.path}/docs/pre-dev/{feature-name}/research.md
  1. Adjust agent dispatch for multi-module projects:

If topology.structure is monorepo or multi-repo:

  • ring:repo-research-analyst: Run ONCE per module path

- Backend: Search {topology.modules.backend.path} for patterns - Frontend: Search {topology.modules.frontend.path} for patterns

  • Combine findings in single research.md with module sections: ### Codebase Research #### Backend Module ({path}) [Agent output for backend] #### Frontend Module ({path}) [Agent output for frontend]
  1. TopologyConfig flows to all subsequent gates:
  • Gate 1 (PRD): Reads topology from research.md frontmatter
  • Gate 2 (Feature Map): Uses topology for domain mapping
  • Gate 7 (Task Breakdown): Tags tasks with target: based on topology

Step 3: Aggregate Research Findings

Output:

  • single-repo/monorepo: docs/pre-dev/{feature-name}/research.md
  • multi-repo: Both {backend.path}/docs/pre-dev/{feature-name}/research.md AND {frontend.path}/docs/pre-dev/{feature-name}/research.md
SectionContent
Metadata (YAML frontmatter)date, feature, research_mode, agents_dispatched, topology
Executive Summary2-3 sentences synthesizing key findings
Research ModeWhy selected, what it means for focus
Codebase ResearchAgent output (file:line references)
Best Practices ResearchAgent output (URLs)
Framework DocumentationAgent output (version constraints)
Product/UX ResearchAgent output (user problem, personas, design constraints)
SynthesisKey patterns to follow (file:line, URL, doc ref); Constraints identified; Prior solutions from docs/solutions/; UX considerations; Open questions for PRD

Step 4: Gate 0 Validation

BLOCKING CHECKLIST:

CheckRequired For
Research mode documentedAll modes
All 4 agents returnedAll modes
research.md createdAll modes
TopologyConfig in frontmatterAll modes (if provided by command)
At least one file:line referencemodification, integration
At least one external URLgreenfield, integration
docs/solutions/ searchedAll modes
Tech stack versions documentedAll modes
Product/UX research documentedAll modes
User problem identifiedAll modes
Synthesis section completeAll modes
Module-specific research (if multi-module)monorepo, multi-repo

If validation fails:

  • Missing agent output → Re-run that agent
  • No codebase patterns (modification) → May need mode change
  • No external docs (greenfield) → Try different search terms
  • No UX research (product-designer) → Re-run product-designer agent

Integration with Pre-Dev Workflow

ring:pre-dev-full (10-gate): Gate 0 Research → Gate 1 PRD (reads research.md) →... → Gate 3 TRD (reads research.md)

ring:pre-dev-feature (5-gate): Gate 0 Research → Gate 1 PRD → Gate 2 TRD → Gate 3 Tasks

Research Document Usage

In Gate 1 (PRD): Reference existing patterns with file:line; cite docs/solutions/; include external URLs; note framework constraints

In Gate 3 (TRD): Reference implementation patterns; use version constraints; cite similar implementations

Anti-Patterns

  1. Skipping research for "simple" features - Even simple features benefit from convention checks
  2. Wrong research mode - Greenfield with heavy codebase research wastes time; modification without codebase research misses patterns
  3. Ignoring docs/solutions/ - Prior solutions are gold; prevents repeating mistakes
  4. Vague references without file:line - "There's a pattern somewhere" is not useful; exact locations enable quick reference

Standards Loading (MANDATORY)

This skill is a research/analysis skill and does NOT require WebFetch of language-specific standards.

Purpose: This skill gathers research BEFORE technical decisions are made. Standards are loaded in TRD (Gate 3) after research establishes context.

However, research agents dispatched by this skill (e.g., ring:framework-docs-researcher) MUST use WebSearch and Context7 to gather current best practices.


Blocker Criteria - STOP and Report

ConditionActionSeverity
Research mode cannot be determinedSTOP and ask user for clarificationCRITICAL
All 4 research agents failed to return resultsSTOP and investigate connectivity/availabilityCRITICAL
No docs/solutions/ directory exists to searchContinue but document as limitationMEDIUM
TopologyConfig provided but invalidSTOP and request valid topologyHIGH
research.md cannot be created (permissions)STOP and report filesystem errorCRITICAL

Cannot Be Overridden

These requirements are NON-NEGOTIABLE:

  • MUST dispatch all 4 research agents in parallel
  • MUST determine research mode BEFORE dispatching agents
  • MUST include product/UX research via ring:product-designer
  • MUST aggregate findings into research.md with file:line references
  • MUST search docs/solutions/ for prior solutions
  • MUST document tech stack versions from manifests
  • CANNOT skip synthesis section

Severity Calibration

SeverityDefinitionExample
CRITICALResearch cannot proceedAll research agents failed, no mode determined
HIGHResearch incomplete but usableMissing agent output (1 of 4)
MEDIUMResearch quality degradedNo file:line references in codebase research
LOWMinor gaps in documentationMissing one external URL reference

Pressure Resistance

User SaysYour Response
"Skip research, just write the PRD""Cannot proceed. Research prevents reinventing existing patterns. Gate 0 takes 30 minutes, saves hours of rework."
"We already researched this manually""Cannot skip automated research. Manual research may miss codebase patterns. I'll run systematic analysis now."
"Only run codebase research, skip external""Cannot skip agents. Mode determines focus, not exclusion. Greenfield needs external research; modification needs both."
"Product research isn't needed, we know users""Cannot skip product-designer. User assumptions need validation. UX research prevents building wrong features."
"docs/solutions/ is empty, skip that check""Cannot skip the check. Empty result is documented. Future features benefit from knowing what doesn't exist."

Anti-Rationalization

RationalizationWhy It's WRONGRequired Action
"Feature is simple, research is overkill"Simple features have conventions too. Research validates simplicityRun all 4 agents regardless of perceived simplicity
"I know the codebase, skip repo research"Personal knowledge isn't documented or sharedRun ring:repo-research-analyst anyway
"External best practices don't apply to us"Best practices exist for reasons. Evaluate, don't dismissRun ring:best-practices-researcher
"Framework docs are the same as last time"Versions change. Patterns evolve. Check current stateRun ring:framework-docs-researcher
"User research slows us down"Building wrong features slows down 100x moreRun ring:product-designer in ux-research mode
"Synthesis is just summary, I can skip it"Synthesis extracts actionable insights. Summary isn't enoughComplete synthesis section with patterns/constraints

When This Skill Is Not Needed

  • Trivial changes that don't need planning (< 1 hour work)
  • research.md already exists and is recent (< 7 days old)
  • Bug fixes with known root cause requiring no design research
  • Documentation-only updates
  • Pure refactoring with no new functionality
  • Configuration changes requiring no architectural research

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

36.74%
按下载量换算104

Claude

29.96%
按下载量换算84

Cursor

16.58%
按下载量换算47

Gemini CLI

8.32%
按下载量换算23

安全审计

暂无安全审计结果可展示。

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。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills