AgentSkillsCN

automation-orchestrator

统筹项目自动化脚本，用于设置、验证与部署。适用于用户希望运行或修改仓库自动化时使用，例如创建 .claude/skills 目录、验证技能、通过 Git 部署技能，或借助辅助脚本检验技术栈的选择。

SKILL.md
--- frontmatter
name: automation-orchestrator
description: >
  Orchestrate project automation scripts for setup, validation, and deployment.
  Use this skill when the user wants to run or modify repo automation, such as
  creating the .claude/skills directory, validating skills, deploying them via git,
  or checking tech stack choices with helper scripts.
allowed-tools: Bash

Automation Orchestrator

Overview

This skill manages and improves project automation scripts in this repository.
Use it whenever the workflow involves running or editing Bash/Python scripts that automate setup, validation, deployment, or routine maintenance tasks.

The primary scripts it coordinates are:

  • •./scripts/claude-skills/setup-claude-skills.sh – Create .claude/skills/ directory structure
  • •./scripts/claude-skills/validate-claude-skills.sh – Verify skills are correctly set up
  • •./scripts/claude-skills/deploy-claude-skills.sh – Commit and push .claude/ to git
  • •./scripts/claude-skills/tech-stack-validator.sh – Validate proposed tech stack choices
  • •./scripts/claude-skills/setup_claude_skills.py – Python variant of the setup script

When to Use This Skill

Trigger this skill when the user says things like:

  • •"Set up Claude skills in this repo"
  • •"Validate the skills setup"
  • •"Deploy the skills to git"
  • •"Run the automation scripts"
  • •"Improve or refactor our setup scripts"
  • •"Add a new script to this automation suite"

If the user is manually creating directories or copying files that scripts already handle, prefer guiding them to use or update the automation instead of duplicating work.

Core Principles

  1. •

    Use existing scripts first
    Prefer calling or improving existing automation (setup, validate, deploy) before writing new one‑off commands.

  2. •

    Idempotent operations
    Scripts should be safe to re‑run without breaking the repo. Avoid instructions that leave the system in a half‑configured state.

  3. •

    Minimal assumptions
    Assume a standard POSIX shell environment and git are available, but do not assume any other tools unless clearly documented.

  4. •

    Explicit side effects
    When suggesting changes, always state what files or git state will be modified.

Standard Workflows

1. Initial Skills Setup

When the user wants to initialize skills for the project:

  1. •

    Confirm they are in the repo root.

  2. •

    Instruct them to run:

    bash
    chmod +x scripts/claude-skills/setup-claude-skills.sh
./scripts/claude-skills/setup-claude-skills.sh
  3. •

    Remind them that SKILL.md files are already in place in each skill folder:

    • •.claude/skills/tech-stack-selector/SKILL.md
    • •.claude/skills/architecture-decisions/SKILL.md
    • •.claude/skills/code-standards-enforcer/SKILL.md
    • •.claude/skills/ci-cd-pipeline-builder/SKILL.md
    • •.claude/skills/agile-executor/SKILL.md
    • •.claude/skills/project-risk-identifier/SKILL.md
    • •.claude/skills/automation-orchestrator/SKILL.md
  4. •

    After setup, run validation (see next workflow).

2. Validate Skills Setup

When the user wants to check that everything is wired correctly:

  1. •

    Ensure the repository has a .claude/skills/ directory.

  2. •

    Instruct:

    bash
    chmod +x scripts/claude-skills/validate-claude-skills.sh
./scripts/claude-skills/validate-claude-skills.sh
  3. •

    Interpret the output:

    • •If any skill is "Missing: SKILL.md" or missing references/ or scripts/, tell the user exactly which directories/files to create or copy.
    • •If all checks pass, confirm that skills are ready to be detected by Claude.

3. Deploy Skills to Git

When the user wants to commit and push .claude/:

  1. •

    Confirm they are okay committing .claude/ to the current repo.

  2. •

    Instruct:

    bash
    chmod +x scripts/claude-skills/deploy-claude-skills.sh
./scripts/claude-skills/deploy-claude-skills.sh
  3. •

    If the script reports "No remote configured", explain how to add one:

    bash
    git remote add origin <your-repo-url>
git push -u origin main
  4. •

    Confirm that team members can now git pull and immediately use the skills.

4. Validate a Proposed Tech Stack

When the user proposes a stack and wants a quick sanity check:

  1. •

    Ask for four items: LANGUAGE, FRAMEWORK, DATABASE, HOSTING.

  2. •

    Map them into the validator script call, for example:

    bash
    chmod +x scripts/claude-skills/tech-stack-validator.sh
./scripts/claude-skills/tech-stack-validator.sh "TypeScript" "Next.js" "PostgreSQL" "Vercel"
  3. •

    Use the script's output as a starting point, then add higher‑level architectural guidance if needed (e.g., scaling, team skill fit).

5. Add a New Automation Script

When the user wants a new script, like create-new-skill.sh:

  1. •

    Propose a filename and high‑level behavior.

  2. •

    Generate a minimal, readable script that:

    • •Accepts arguments (e.g., skill name)
    • •Creates needed directories under .claude/skills/
    • •Writes a starter SKILL.md template
  3. •

    Remind the user to:

    bash
    chmod +x <script-name>.sh
  4. •

    Suggest adding a short usage example in the repo README.

Editing and Improving Scripts

When asked to change a script:

  1. •Read the existing script logic first rather than replacing it blindly.
  2. •Preserve:
    • •Error handling (set -e), if present
    • •Helpful log messages
    • •Exit codes
  3. •Suggest incremental improvements:
    • •Input validation (check required args)
    • •Safer operations (e.g., mkdir -p instead of mkdir)
    • •Clearer messages for success/failure

If a change might be breaking, explicitly warn the user and propose backing up the original script.

Examples

Example 1: First‑Time Setup

User: "Set up the Claude skills folder structure for this repo."
You (using this skill):

  1. •

    Confirm they are in the project root.

  2. •

    Suggest:

    bash
    chmod +x scripts/claude-skills/setup-claude-skills.sh
./scripts/claude-skills/setup-claude-skills.sh
  3. •

    Confirm that SKILL.md files are already in place.

  4. •

    Then run:

    bash
    ./scripts/claude-skills/validate-claude-skills.sh

Example 2: Deploy Skills

User: "Push these skills so my team can use them."

  1. •

    Ask if .claude/ should be part of the main repo.

  2. •

    If yes, instruct:

    bash
    chmod +x scripts/claude-skills/deploy-claude-skills.sh
./scripts/claude-skills/deploy-claude-skills.sh
  3. •

    If there is no git remote, show how to add one and push.

Example 3: Check a Stack

User: "Validate TypeScript + Next.js + PostgreSQL + Vercel as a stack."

  1. •

    Map to script call:

    bash
    ./scripts/claude-skills/tech-stack-validator.sh "TypeScript" "Next.js" "PostgreSQL" "Vercel"
  2. •

    Summarize the output and add architectural advice on top (e.g., where this stack works best, scaling concerns).

Troubleshooting

  • •

    Problem: ./scripts/claude-skills/setup-claude-skills.sh: Permission denied
    Action: Instruct:

    bash
    chmod +x scripts/claude-skills/setup-claude-skills.sh
./scripts/claude-skills/setup-claude-skills.sh
  • •

    Problem: Validation script reports missing SKILL.md
    Action: Tell the user which specific path is missing, and that they should copy the corresponding SKILL.md there.

  • •

    Problem: Git commit in deployment script fails
    Action: Ask for the exact error message. Check:

    • •Repo is initialized (.git/ exists)
    • •User name/email configured (git config user.name, git config user.email)
    • •Remote is set for pushing, if desired.

Use this skill to keep automation centralized, repeatable, and safe, rather than scattering ad‑hoc commands across the project.