AgentSkillsCN

skill-creator

为 Cycle Navigator 项目生成新代理技能的指导原则。适用于用户希望为特定组件或工作流创建新的 SKILL.md 文件时使用。

SKILL.md
--- frontmatter
name: skill-creator
description: Guidelines for generating new Agent Skills for the Cycle Navigator project. Use this when the user wants to create a new SKILL.md file for a specific component or workflow.

Skill Creator

Overview

This skill provides structured guidance for creating new Agent Skills that are consistent with the Cycle Navigator project's technical architecture, developer setup, and best practices.

Required Frontmatter

Every generated skill must start with this exact YAML structure:

yaml
---
name: lowercase-with-hyphens
description: A clear "Trigger" sentence (e.g., "Use this when...") so Copilot knows when to activate it.
---
  • •name: Must be lowercase with hyphens, no spaces. Examples: skill-creator, macro-data-fetcher, celery-task-builder
  • •description: Start with "Use this when..." to provide clear trigger conditions for skill discovery.

Core Project Context to Include

Always reference these project-specific facts if relevant to the new skill:

  • •Environment: Fedora Linux using Podman/Docker Compose
  • •Backend: FastAPI (Python 3.11+), located in backend/ directory
  • •Database: PostgreSQL 16 + TimescaleDB (⚠️ hypertables are irreversible - always verify before creating)
  • •Workers: Celery using the backend.celery_app module
  • •Frontend: Next.js + TypeScript, located in web/ directory
  • •Package Management: pyproject.toml (Python), package.json (Node.js)

Formatting Standards

  • •Use Markdown headers (##, ###) for clear organization
  • •Provide example commands using the project's preferred tools:
    • •podman-compose instead of generic docker
    • •pip with pyproject.toml conventions
    • •Terminal code blocks with clear explanations
  • •Include a Safety/Verification section for any skill involving:
    • •Database schema changes
    • •Code deployments
    • •Celery task creation
    • •Environment modifications

Folder Structure

Place new skills in:

code
.github/skills/<skill-name>/
├── SKILL.md              # Main instructions (required)
├── examples.md           # Usage examples (optional but recommended)
├── scripts/              # Utility scripts (optional)
│   └── utility.py

Skill Template

Use this as a starting point for new skills:

markdown
---
name: your-skill-name
description: Brief description of functionality. Use this when [specific trigger or use case].
---

# Your Skill Name

## Overview
[1-2 sentence description of what this skill enables]

## When to Use This Skill
[Explain the trigger conditions clearly]

## Instructions
[Step-by-step guidance for Claude/Copilot to follow]

### Setup
[Any preliminary setup or environment checks]

### Implementation
[Detailed steps]

### Verification
[How to verify the task was completed successfully]

## Examples

### Example 1: [Specific scenario]
[Walkthrough of using the skill in this scenario]

## Safety/Verification
[Important warnings, irreversible operations, or validation steps]

## Troubleshooting
[Common issues and solutions specific to this project]

Version Requirements

Ensure any skill you create acknowledges these project constraints:

  • •Python: 3.11 or higher
  • •PostgreSQL: 16
  • •Node.js: Check web/package.json for current LTS version
  • •Podman/Docker: Latest stable versions

Testing Your Skill

Before finalizing a new skill:

  1. •Syntax check: Validate YAML frontmatter
  2. •Context test: Ensure instructions reference correct file paths and modules
  3. •Tool verification: Confirm any tools or scripts mentioned exist in the project
  4. •Trigger clarity: Verify the description clearly indicates when Copilot should activate this skill

Examples of Well-Structured Skills

Reference these projects as examples:

  • •Dev Agent Skills (/fvadicamo/dev-agent-skills): Git, GitHub, Conventional Commits
  • •Supabase Agent Skills (/supabase/agent-skills): Database and API interactions
  • •Vercel Agent Skills (/vercel-labs/agent-skills): Deployment and infrastructure

Tips for Success

  • •Be specific: Use exact module names (backend.celery_app, not just celery)
  • •Include context: Always mention whether a task affects frontend, backend, or both
  • •Warn about irreversibility: TimescaleDB hypertables, migrations, and schema changes need explicit warnings
  • •Reference docs: Link to TECHNICAL_ARCHITECTURE.md and DEVELOPER_SETUP.md when relevant
  • •Make it actionable: Skills should enable users to complete tasks without leaving VS Code

Skills Location

Store all new skills in: .github/skills/<skill-name>/SKILL.md

This follows the Agent Skills Standard and ensures consistent discovery by Claude Code and GitHub Copilot.