plan-based-skill-builder

📁 sayujks0071/antidhan 📅 6 days ago
4
总安装量
4
周安装量
#48005
全站排名
安装命令
npx skills add https://github.com/sayujks0071/antidhan --skill plan-based-skill-builder

Agent 安装分布

opencode 4
claude-code 4
codex 4
mcpjam 3
openhands 3
zencoder 3

Skill 文档

Plan-Based Skill Builder

Quick Start

When building a skill from a plan document:

  1. Analyze the plan document structure and purpose
  2. Extract workflows, checklists, and key instructions
  3. Identify trigger scenarios and use cases
  4. Create a structured SKILL.md following best practices
  5. Generate supporting files if needed (reference.md, examples.md)

Plan Document Types

Restart Plans

  • Focus on restart procedures, verification steps, health checks
  • Extract restart workflows, validation commands, troubleshooting steps
  • Create skills for restart operations, system recovery, health monitoring

Deployment Plans

  • Focus on deployment steps, configuration, validation
  • Extract deployment workflows, configuration templates, rollback procedures
  • Create skills for deployment operations, configuration management

Prioritization Plans

  • Focus on ranking, scoring, decision-making frameworks
  • Extract scoring rubrics, ranking algorithms, decision criteria
  • Create skills for prioritization, ranking, selection workflows

Retest Plans

  • Focus on testing procedures, validation steps, success criteria
  • Extract test workflows, validation commands, acceptance criteria
  • Create skills for testing, validation, quality assurance

Migration Plans

  • Focus on migration steps, data transformation, validation
  • Extract migration workflows, transformation rules, verification steps
  • Create skills for migration operations, data transformation

Skill Creation Workflow

Step 1: Analyze Plan Document

Read the plan document and identify:

Purpose & Scope:

  • What is the main goal of this plan?
  • What domain/area does it cover?
  • What problems does it solve?

Key Workflows:

  • What are the main steps or procedures?
  • Are there checklists or sequential processes?
  • What decision points exist?

Trigger Scenarios:

  • When would someone need this skill?
  • What keywords or situations trigger its use?
  • What related tasks would benefit from this?

Key Instructions:

  • What specific commands or procedures are documented?
  • What patterns or templates are provided?
  • What validation or verification steps exist?

Supporting Information:

  • Are there examples or reference materials?
  • Are there scripts or utilities mentioned?
  • Are there integration points with other systems?

Step 2: Extract Skill Components

Name:

  • Use lowercase, hyphens, max 64 chars
  • Reflect the plan’s purpose (e.g., restart-operations, deployment-workflow)
  • Be specific and descriptive

Description:

  • Write in third person
  • Include WHAT (capabilities) and WHEN (trigger scenarios)
  • Include key terms for discovery
  • Max 1024 chars

Instructions:

  • Convert plan steps into actionable instructions
  • Preserve workflows and checklists
  • Include command examples and code snippets
  • Add validation steps and error handling

Examples:

  • Extract concrete examples from the plan
  • Include command-line examples
  • Show expected outputs or results

Integration Points:

  • Identify related skills or systems
  • Document dependencies
  • Reference related workflows

Step 3: Structure the Skill

Follow the standard SKILL.md structure:

---
name: skill-name
description: [Third-person description with triggers]
---

# Skill Name

## Quick Start
[Essential instructions]

## Workflow
[Step-by-step process]

## Key Components
[Important concepts, commands, patterns]

## Examples
[Concrete examples]

## Integration Points
[Related skills, systems, workflows]

## Troubleshooting
[Common issues and solutions]

Step 4: Create Supporting Files (If Needed)

reference.md:

  • Detailed API documentation
  • Complete command reference
  • Advanced configurations

examples.md:

  • Extended examples
  • Use case scenarios
  • Edge cases

scripts/ (if needed):

  • Utility scripts mentioned in plan
  • Validation scripts
  • Helper tools

Plan Analysis Patterns

Restart Plan Pattern

Common Sections:

  • Restart procedures (stop → wait → start)
  • Health checks and verification
  • Troubleshooting steps
  • Rollback procedures

Skill Extraction:

## Restart Workflow

1. Stop services: [commands]
2. Wait for cleanup: [duration]
3. Start services: [commands]
4. Verify health: [checks]
5. Validate functionality: [tests]

Deployment Plan Pattern

Common Sections:

  • Pre-deployment checks
  • Deployment steps
  • Post-deployment validation
  • Rollback procedures

Skill Extraction:

## Deployment Process

### Pre-Deployment
- [ ] Checklist item 1
- [ ] Checklist item 2

### Deployment
1. Step 1: [command]
2. Step 2: [command]

### Post-Deployment
- Verify: [check]
- Validate: [test]

Prioritization Plan Pattern

Common Sections:

  • Scoring framework
  • Ranking algorithm
  • Decision criteria
  • Action recommendations

Skill Extraction:

## Scoring Framework

### Factor 1: [Name]
- **5**: [Criteria]
- **4**: [Criteria]
- **3**: [Criteria]

### Ranking Process
1. Score each item
2. Calculate composite
3. Rank by score
4. Categorize by action

Example: Building from Restart Plan

Input Plan Document:

# System Restart Plan

## Restart Procedure
1. Stop API: `kill -TERM <PID>`
2. Wait 10 seconds
3. Start API: `make start-paper`
4. Verify: `curl :8000/ready`

Generated Skill:

---
name: system-restart
description: Restart trading systems with proper sequencing and health verification. Use when restarting APIs, services, or when system recovery is needed.
---

# System Restart

## Quick Start

Restart sequence:
1. Stop services gracefully
2. Wait for cleanup
3. Start services
4. Verify health

## Restart Workflow

### Step 1: Stop Services
```bash
# Find and stop API
lsof -nP -iTCP:8000 | awk '/LISTEN/ {print $2}' | xargs kill -TERM

Step 2: Wait for Cleanup

sleep 10  # Allow graceful shutdown

Step 3: Start Services

make start-paper  # or make live

Step 4: Verify Health

curl :8000/ready | jq  # Should return 200
curl :8000/metrics | grep '^trader_is_leader'  # Should be 1

## Best Practices

### 1. Preserve Plan Structure
- Keep workflows intact
- Maintain checklists
- Preserve command sequences

### 2. Enhance with Context
- Add error handling not in plan
- Include troubleshooting steps
- Add validation checks

### 3. Make It Discoverable
- Include trigger keywords in description
- Add related terms
- Reference common use cases

### 4. Keep It Concise
- Extract essential information
- Remove redundant explanations
- Focus on actionable content

### 5. Add Integration Points
- Reference related skills
- Link to documentation
- Connect to related workflows

## Common Extraction Patterns

### Commands → Instructions
```markdown
# Plan:
```bash
make start-paper

Skill:

Start System

make start-paper

Wait 10 seconds, then verify with curl :8000/ready


### Checklists → Workflows
```markdown
# Plan:
- [ ] Step 1
- [ ] Step 2

# Skill:
## Deployment Checklist
1. Step 1: [details]
2. Step 2: [details]

Procedures → Step-by-Step

# Plan:
1. Do X
2. Then Y
3. Finally Z

# Skill:
## Procedure
### Step 1: Do X
[Details and commands]

### Step 2: Then Y
[Details and commands]

### Step 3: Finally Z
[Details and commands]

Validation Checklist

After creating a skill from a plan, verify:

  • Skill name is lowercase, hyphenated, max 64 chars
  • Description is third-person, includes triggers
  • Description includes WHAT and WHEN
  • Instructions are actionable and clear
  • Workflows preserve plan structure
  • Examples are concrete and useful
  • Integration points are documented
  • SKILL.md is under 500 lines
  • Supporting files are one level deep
  • Commands are tested and accurate

Related Resources

  • Skill creation guide: create-skill skill
  • Plan documents: *PLAN*.md files in root
  • Existing skills: .cursor/skills/ directory
  • Agent files: .cursor/agents/ directory

Troubleshooting

Plan Too Vague

  • Look for implicit workflows
  • Extract common patterns
  • Add general guidance

Plan Too Specific

  • Generalize to broader use cases
  • Extract reusable patterns
  • Add configuration options

Missing Information

  • Infer from context
  • Add placeholder sections
  • Document assumptions

Conflicting Information

  • Use most recent or authoritative source
  • Document conflicts
  • Provide alternatives

Always create skills that are discoverable, actionable, and aligned with the plan’s intent while following Cursor skill best practices.