Token导航 LogoToken导航TokenDH.com
研究检索权限需确认github未标认证来源可访问clear审计未展示

increment-work-router增量工作路由器

Agent Skill

increment-work-router 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

420

周安装

17

GitHub Stars

127

下载量

132
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

3

许可证

MIT

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:increment-work-router(增量工作路由器)
来源仓库:https://github.com/anton-abyzov/specweave
仓库路径:skills/increment-work-router
安装命令:
npx skills add https://github.com/anton-abyzov/specweave --skill increment-work-router
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。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/anton-abyzov/specweave --skill increment-work-router

简介

increment-work-router 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中快速定位候选结果。

  • 适用于关键词搜索、任务场景分析和来源线索整理等研究检索场景。
  • 通过 npx skills add 命令从 GitHub 仓库安装并使用该技能。
  • 安装前需确认权限范围、维护状态及是否触发联网或文件操作。
  • 建议结合原始 README 文档进一步核验具体用法和功能边界。

SKILL.md

Increment Work Router - Smart Work Continuation

Purpose

The increment-work-router skill is an intelligent work continuation system that:

  • ✅ Detects implementation/continuation intent (not just new projects)
  • ✅ Checks for active increments automatically
  • ✅ Routes to existing increment OR creates new one
  • ✅ Bridges the gap between planning and execution

Key Insight: When user says "implement X" or "complete Y", they expect the system to handle routing intelligently without manual commands.

When to Activate

This skill activates when user expresses implementation/continuation intent:

High-Confidence Triggers (Auto-Route)

Action Verbs (with specific target):

  • "Implement [feature]"
  • "Complete [task]"
  • "Build [component]"
  • "Add [functionality]"
  • "Develop [module]"
  • "Create [feature]"

Continuation Phrases:

  • "Work on [feature]"
  • "Continue [increment]"
  • "Resume [work]"
  • "Finish [task]"
  • "Let's implement [X]"
  • "Let's build [Y]"
  • "Start working on [Z]"

Bug/Fix Intent:

  • "Fix [issue]"
  • "Resolve [bug]"
  • "Address [problem]"

Medium-Confidence Triggers (Clarify)

Vague Intent (needs target clarification):

  • "Let's continue"
  • "Keep working"
  • "What's next?"
  • "Continue where we left off"

Don't Activate For

Planning/Discussion (let other skills handle):

  • "What should we build?" → increment-planner
  • "How does X work?" → Regular conversation
  • "Should we use Y?" → Technical discussion
  • "Plan a new feature" → increment-planner

Already in Workflow:

  • User is already executing /sw:do
  • Increment planning is in progress
  • Another skill is handling the request

Core Algorithm

Step 1: Detect Intent

Scan user message for implementation keywords:

Action verbs: implement, complete, build, add, develop, create
Continuation: work on, continue, resume, finish, start
Bug/Fix: fix, resolve, address

Calculate confidence:

  • High (>80%): Action verb + specific target ("implement user auth")
  • Medium (50-80%): Action verb only ("let's continue")
  • Low (<50%): No clear intent

Step 2: Check Active Increments

Read increment state:

# Find all increments with status = "active"
find .specweave/increments -name "metadata.json" -type f \
  -exec jq -r 'select(.status == "active") | .id' {} \;

Three scenarios:

  1. One active increment → Check relevance and route
  2. Multiple active increments → Ask user to select
  3. No active increments → Suggest creating new one

Step 3: Check Relevance (if active increment exists)

Relatedness Analysis: Compare user's request against active increment:

  • Check increment title/description
  • Check task list (tasks.md)
  • Check spec.md for related features

Scoring:

  • High match (>70%): Same feature area → Auto-resume
  • Medium match (40-70%): Related area → Ask confirmation
  • Low match (<40%): Unrelated → Suggest new increment

Step 4: Route Intelligently

Scenario A: High relevance to active increment

User: "Implement user authentication"
Active: 0031-user-authentication-system

✅ High match → Auto-resume:

"I see you have an active increment for user authentication (0031).
Resuming work on that increment now with /sw:do..."

[Automatically invokes /sw:do]

Scenario B: Unrelated to active increment

User: "Implement payment processing"
Active: 0031-user-authentication-system

⚠️  Low match → Ask:

"I see you have an active increment (0031-user-authentication-system).

Your request for payment processing seems unrelated. Would you like to:
1. **Create new increment** for payment processing (recommended)
2. **Continue current increment** (0031) and add payment to it
3. **Pause current** and create new increment

What would you prefer?"

Scenario C: No active increment

User: "Implement user authentication"
Active: None

→ Suggest new increment:

"No active increment found. Let's create one for user authentication.

I'm going to create a SpecWeave increment with:
- Specification (WHAT & WHY)
- Technical plan (HOW)
- Task breakdown (STEPS with embedded tests)

Creating increment now with /sw:increment..."

[Automatically invokes /sw:increment "user authentication"]

Scenario D: Multiple active increments

User: "Continue working"
Active: 0031-user-auth, 0032-payment-system

→ Ask to clarify:

"You have 2 active increments:
1. **0031-user-authentication-system** (3/10 tasks complete)
2. **0032-payment-processing** (1/8 tasks complete)

Which one would you like to work on?"

Scenario E: Vague continuation intent

User: "Let's continue"
Active: 0031-user-authentication-system

✅ Auto-resume (only one active):

"Resuming increment 0031-user-authentication-system...
Progress: 3/10 tasks complete (30%)

Next task: T-004 Implement password hashing

[Automatically invokes /sw:do]"

Integration Architecture

Workflow:

User: "Implement [feature]"
        ↓
increment-work-router (detects intent)
        ↓
[Check .specweave/increments/*/metadata.json]
        ↓
┌───────────────────┬──────────────────┬────────────────┐
│ Active increment  │ Multiple active  │ No active      │
│ (relevant)        │ (ambiguous)      │ (none)         │
└─────┬─────────────┴──────┬───────────┴──────┬─────────┘
      │                    │                  │
      ↓                    ↓                  ↓
/sw:do        Ask to select    /sw:increment
(auto-resume)        (clarify)        (create new)

Calls:

  • /sw:do - Resume active increment
  • /sw:increment - Create new increment
  • /sw:status - Check increment state (if needed)
  • /sw:tdd-cycle - TDD workflow (when TDD mode enabled)

Called By:

  • Automatically when implementation intent detected
  • Works alongside increment-planner (planning) and detector (context checking)

TDD-Aware Routing (CRITICAL)

When routing to an active increment, check TDD mode first:

# Check if increment uses TDD
CONFIG_PATH=".specweave/config.json"
METADATA_PATH=".specweave/increments/<id>/metadata.json"

# Check global config
TDD_MODE=$(cat "$CONFIG_PATH" | jq -r '.testing.defaultTestMode // "test-after"')

# Check increment-specific override
INCREMENT_TDD=$(cat "$METADATA_PATH" | jq -r '.testMode // ""')
[[ -n "$INCREMENT_TDD" ]] && TDD_MODE="$INCREMENT_TDD"

If TDD mode is enabled, modify routing behavior:

ScenarioWithout TDDWith TDD
"Implement X" (new feature)/sw:do→ Suggest /sw:tdd-cycle first
"Let's continue"/sw:do→ Show TDD phase reminder + /sw:do
"Add test for X"/sw:do→ Confirm starting RED phase
"Fix the implementation"/sw:do→ Check if GREEN phase complete

TDD-aware resume output:

✅ Resuming increment 0031-user-authentication-system...

🔴 TDD MODE ACTIVE

Current TDD Status:
├─ T-001: [RED] Write login test ✅ completed
├─ T-002: [GREEN] Implement login ⏳ in progress
└─ T-003: [REFACTOR] Clean up login ⏸️ blocked (waiting for GREEN)

Current Phase: 🟢 GREEN - Making test pass

💡 You're in the GREEN phase. Implement just enough to make T-001's test pass.
   After GREEN completes, you can proceed to REFACTOR.

[Proceeding with /sw:do...]

TDD workflow suggestion (for new work):

User: "Implement user registration"

🔴 TDD MODE DETECTED

This increment uses Test-Driven Development.

For new features, I recommend using the TDD workflow:
1. /sw:tdd-red "user registration" - Write failing test first
2. /sw:tdd-green - Implement to pass the test
3. /sw:tdd-refactor - Clean up the code

Would you like to:
1. Start TDD cycle (/sw:tdd-cycle) - Recommended
2. Continue with regular /sw:do - Skip TDD guidance

[1/2]:

Decision Matrix

User IntentActive IncrementsRelevanceAction
"Implement auth"1 (auth-related)High (>70%)Auto /sw:do
"Implement auth"1 (unrelated)Low (<40%)Ask: New or add to current?
"Implement auth"0N/AAuto /sw:increment
"Implement auth"2+N/AAsk which increment
"Let's continue"1N/AAuto /sw:do
"Let's continue"2+N/AAsk which increment
"Let's continue"0N/A"No active increment. What should we build?"

Relevance Matching Logic

How to determine if request relates to active increment:

function calculateRelevance(userRequest: string, increment: Increment): number {
  let score = 0;

  // 1. Title/ID match (strongest signal)
  if (increment.title.toLowerCase().includes(extractKeywords(userRequest))) {
    score += 40;
  }

  // 2. Tasks match (check tasks.md)
  const tasks = readTasksFile(increment);
  const matchingTasks = tasks.filter(t =>
    t.title.toLowerCase().includes(extractKeywords(userRequest))
  );
  score += (matchingTasks.length / tasks.length) * 30;

  // 3. Spec match (check spec.md user stories)
  const spec = readSpecFile(increment);
  if (spec.includes(extractKeywords(userRequest))) {
    score += 20;
  }

  // 4. Type match (feature vs bug vs hotfix)
  const requestType = detectType(userRequest);
  if (requestType === increment.type) {
    score += 10;
  }

  return score; // 0-100
}

Example:

User: "Implement JWT token refresh"
Active: 0031-user-authentication-system

Checks:
✅ Title contains "authentication" (user mentions "JWT token") → +40 points
✅ Tasks.md has "T-005: Implement JWT refresh logic" → +30 points
✅ Spec.md mentions "token management" → +20 points
✅ Both are "feature" type → +10 points

Total: 100/100 → High relevance → Auto-resume

Examples

Example 1: Auto-Resume (High Relevance)

Context:

Active: 0031-user-authentication-system
Progress: 3/10 tasks complete

User:

Implement JWT token refresh

increment-work-router:

✅ Intent detected: "implement" + specific target
✅ Active increment found: 0031-user-authentication-system
✅ Relevance check: 95% match (authentication-related)

→ AUTO-RESUME

"I see you have an active increment for user authentication (0031).
Task T-005 'Implement JWT token refresh' is already in your plan.

Resuming increment 0031 now with /sw:do...

[Automatically invokes /sw:do]

Example 2: Create New (No Active Increment)

Context:

Active: None

User:

Implement payment processing with Stripe

increment-work-router:

✅ Intent detected: "implement" + specific target
❌ No active increment found

→ CREATE NEW

"No active increment found. Let's create one for payment processing.

I'm going to create a SpecWeave increment with:
- Specification (Stripe integration requirements)
- Technical plan (API integration, webhooks, security)
- Task breakdown (implementation steps with tests)

Creating increment now with /sw:increment...

[Automatically invokes /sw:increment "payment processing with Stripe"]

Example 3: Unrelated Request (Ask)

Context:

Active: 0031-user-authentication-system
Progress: 3/10 tasks complete

User:

Implement email notifications

increment-work-router:

✅ Intent detected: "implement" + specific target
✅ Active increment found: 0031-user-authentication-system
⚠️  Relevance check: 15% match (unrelated)

→ ASK FOR CLARIFICATION

"I see you have an active increment (0031-user-authentication-system) with 3/10 tasks complete.

Your request for email notifications seems unrelated. Would you like to:
1. **Create new increment** for email notifications (recommended - keeps work focused)
2. **Add to current increment** (0031) - if notifications are for auth emails
3. **Pause current increment** and start notifications

SpecWeave works best with focused increments. What would you prefer?"

Example 4: Multiple Active (Clarify)

Context:

Active: 0031-user-authentication-system (3/10 tasks)
Active: 0032-payment-processing (1/8 tasks)

User:

Let's continue working

increment-work-router:

✅ Intent detected: "continue" (vague)
✅ Multiple active increments found: 2

→ ASK TO SELECT

"You have 2 active increments:

1. **0031-user-authentication-system**
   - Progress: 3/10 tasks (30%)
   - Next: T-004 Implement password hashing

2. **0032-payment-processing**
   - Progress: 1/8 tasks (12%)
   - Next: T-002 Set up Stripe SDK

Which one would you like to work on?"

[Wait for user selection, then invoke /sw:do {selected-id}]

Example 5: Vague Intent + Single Active (Auto-Resume)

Context:

Active: 0031-user-authentication-system (3/10 tasks)

User:

What's next?

increment-work-router:

✅ Intent detected: "what's next" (continuation)
✅ One active increment: 0031-user-authentication-system

→ AUTO-RESUME

"Resuming increment 0031-user-authentication-system...
Progress: 3/10 tasks complete (30%)

Next task: **T-004 Implement password hashing**

[Automatically invokes /sw:do]"

Opt-Out Mechanism

Users can override auto-routing with explicit instructions:

  • "Don't resume, create new" → Forces new increment
  • "Just discuss first" → Regular conversation
  • "Plan without implementing" → Routes to /sw:increment only
  • "Show me the current state" → Uses /sw:status instead

Success Criteria

  • ✅ Users can say "implement X" and work starts automatically
  • ✅ Smart routing to active increments (no manual /sw:do)
  • ✅ Detects unrelated requests and prevents scope creep
  • ✅ No "which increment?" confusion (auto-handles single active)
  • ✅ Clear choices when ambiguous (multiple active or unrelated)
  • ✅ Seamless integration with existing skills (increment-planner, detector)

Related Skills

  • increment-planner: For creating increment structure (invoked by this skill for new projects)
  • detector: For checking SpecWeave context

Key Distinction:

  • increment-planner = "PLAN this increment" or "I want to BUILD a new product" (planning-level)
  • increment-work-router = "IMPLEMENT this feature/task" (execution-level)

This skill bridges planning → execution by auto-detecting implementation intent.

Project-Specific Learnings

Before starting work, check for project-specific learnings:

# Check if skill memory exists for this skill
cat .specweave/skill-memories/increment-work-router.md 2>/dev/null || echo "No project learnings yet"

Project learnings are automatically captured by the reflection system when corrections or patterns are identified during development. These learnings help you understand project-specific conventions and past decisions.

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

04

需要参考平台分布和安装热度时

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

能力 4

补充不同宿主或平台的使用分布数据

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

Claude Code

27.71%
按下载量换算37

Antigravity

24.61%
按下载量换算32

Cursor

16.84%
按下载量换算22

Gemini CLI

12.36%
按下载量换算16

OpenCode

8.51%
按下载量换算11

Codex

3.24%
按下载量换算4

安全审计

暂无安全审计结果可展示。

权限和风险

权限需确认

当前来源未能明确判断权限范围,默认进入异常复核队列。

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。

来源信息

继续浏览同类 Skills