用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/RunnerQuan/SAFE-Agent --skill workflow-composer命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | workflow-composer |
| description | Chain multiple skills together into automated workflows with conditional logic and parallel execution |
| allowed-tools | ["Read","Write","Bash","Task","TodoWrite"] |
| version | 1.0.0 |
| author | GLINCKER Team |
| license | Apache-2.0 |
| keywords | ["automation","workflow","orchestration","skill-chaining","multi-agent","pipeline"] |
⚡ UNIQUE FEATURE: First skill that lets you chain multiple Claude Code skills together into automated workflows with conditional logic, parallel execution, and error handling. Create complex automation pipelines with simple YAML configs.
Orchestrates multiple skills into powerful automated workflows:
This is the first skill-composition framework for Claude Code:
name: my-workflow
description: What this workflow does
version: 1.0.0
# Input parameters
inputs:
project_path:
type: string
required: true
run_tests:
type: boolean
default: true
# Workflow steps
steps:
# Sequential steps
- name: generate-readme
skill: readme-generator
inputs:
path: ${inputs.project_path}
- name: generate-tests
skill: unit-test-generator
inputs:
target: ${inputs.project_path}/src
when: ${inputs.run_tests}
# Parallel execution
- name: quality-checks
parallel:
- name: security-scan
skill: security-auditor
inputs:
path:
When a user wants to create a workflow:
Interactive Workflow Builder:
Ask user:
- What's the goal of this workflow?
- Which skills should be included?
- Should any skills run in parallel?
- What are the inputs needed?
- What conditions/logic are needed?
Generate Workflow YAML:
Save Workflow:
.claude-workflows/workflow-name.ymlWhen a user wants to run a workflow:
Load Workflow:
Use Read to load workflow YAML from .claude-workflows/
Validate Inputs:
Create Execution Plan:
- Parse workflow steps
- Identify parallel vs sequential
- Build dependency graph
- Prepare Task agents
Execute Workflow:
For each step:
- Check 'when' condition (skip if false)
- If parallel block:
- Launch all Task agents simultaneously
- Wait for all to complete
- If sequential:
- Execute skill via Task agent
- Capture output
- Pass to next step
- Handle errors:
- Retry if configured
- Run on_failure steps
- Stop or continue based on config
Track Progress:
Generate Report:
# Workflow Execution Report
**Workflow**: ${workflow.name}
**Started**: ${start_time}
**Duration**: ${duration}
**Status**: ✅ Success / ⚠️ Partial / ❌ Failed
## Steps
1. ✅ generate-readme (2.3s)
2. ✅ generate-tests (5.1s)
3. ⏭️ fix-issues (skipped - no issues found)
4. ✅ deploy (12.4s)
## Outputs
- readme_path: /project/README.md
- test_coverage: 87%
- deployment_url: https://app.example.com
name: project-setup
description: Initialize new project with README, tests, CI/CD, and docs
steps:
- name: scaffold
skill: project-scaffolder
inputs:
type: ${inputs.project_type}
name: ${inputs.project_name}
- name: generate-files
parallel:
- name: readme
skill: readme-generator
- name: gitignore
skill: gitignore-generator
- name: license
skill: license-picker
inputs:
license_type: MIT
- name: setup-tests
skill: unit-test-generator
inputs:
coverage_target: 80
- name: setup-cicd
Usage:
User: "Set up a new Python project called 'my-api'"
Workflow Composer:
- Prompts for project type, git remote
- Executes all steps
- Shows progress
- Reports results
name: pre-commit-checks
description: Run all quality checks before committing code
steps:
- name: quality-checks
parallel:
- name: lint
skill: code-linter
- name: format
skill: code-formatter
- name: security
skill: security-auditor
- name: tests
skill: test-runner
- name: fix-issues
when: ${steps.quality-checks.lint.issues > 0}
skill: auto-fixer
inputs:
auto_fix: true
- name: verify
when: ${steps.fix-issues.ran}
skill: test-runner
name: deploy-pipeline
description: Complete deployment with tests, build, and monitoring setup
steps:
- name: pre-deploy-checks
parallel:
- skill: test-runner
- skill: security-auditor
- skill: dependency-checker
- name: build
when: ${steps.pre-deploy-checks.all_passed}
skill: build-optimizer
- name: deploy
skill: deploy-orchestrator
inputs:
environment: ${inputs.environment}
strategy: blue-green
retry:
max_attempts: 3
- name: post-deploy
parallel:
- skill: health-checker
- skill: performance-tester
variables:
project_name: my-app
version: 1.0.0
deploy_env: production
steps:
- name: deploy
inputs:
version: ${variables.version}
env: ${variables.deploy_env}
- name: deploy-prod
when: |
${inputs.environment == 'production'} &&
${steps.tests.coverage >= 80} &&
${steps.security.issues == 0}
- name: process-files
for_each: ${inputs.files}
skill: file-processor
inputs:
file: ${item}
- name: critical-step
skill: deployment
on_failure:
- skill: rollback
- skill: notify-team
continue_on_error: false
Share workflows with the community:
# Publish workflow
claude workflow publish my-workflow.yml
# Install community workflow
claude workflow install feature-complete-setup
# List available workflows
claude workflow browse
This skill works with ANY other skill in the marketplace!
Popular combinations:
This is a cornerstone skill for the marketplace. Help us improve:
Apache License 2.0 - See LICENSE
GLINCKER Team
🌟 WORLD'S FIRST skill composition framework for Claude Code!