Back to skills
SKILL.md
Dispatch Coding Agents From Github Issue Labels And Return Pull Requests With Patchwork Agents
ASecurityUse issue labels as a lightweight dispatch layer that fans repository work out to Claude Code, Codex, or Aider workers and brings back PRs.
- 36 stars
- 0 votes
- 0 copies
- 3 views
- Added June 8, 2026
Works with
Security analysis
100/100npx -y skills add agentskillexchange/skills --skill dispatch-coding-agents-from-github-issue-labels-and-return-pull-requests-with-patchwork-agents --agent claude-codeAre you the author of Dispatch Coding Agents From Github Issue Labels And Return Pull Requests With Patchwork Agents?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/agentskillexchange-dispatch-coding-agents-from-github-issue-labels-an)---
name: "Dispatch coding agents from GitHub issue labels and return pull requests with Patchwork Agents"
slug: "dispatch-coding-agents-from-github-issue-labels-and-return-pull-requests-with-patchwork-agents"
description: "Use issue labels as a lightweight dispatch layer that fans repository work out to Claude Code, Codex, or Aider workers and brings back PRs."
github_stars: 0
verification: "security_reviewed"
source: "https://github.com/hey-intent/patchwork-agents"
author: "hey-intent"
publisher_type: "organization"
category: "CI/CD Integrations"
framework: "Multi-Framework"
tool_ecosystem:
github_repo: "hey-intent/patchwork-agents"
github_stars: 0
---
# Dispatch coding agents from GitHub issue labels and return pull requests with Patchwork Agents
Use issue labels as a lightweight dispatch layer that fans repository work out to Claude Code, Codex, or Aider workers and brings back PRs.
## Prerequisites
GitHub App or webhook setup, Kubernetes or k3s environment, supported coding-agent credentials, repository access, Patchwork Agents deployment
## Installation
Use the upstream install or setup path that matches your environment:
- docker build -f images/orchestrator/Dockerfile -t ghcr.io/<your-org>/orchestrator:latest .
- docker save ghcr.io/<your-org>/orchestrator:latest | sudo k3s ctr images import -
- docker build -f images/worker-claude/Dockerfile -t worker-claude:latest .
- docker save worker-claude:latest | sudo k3s ctr images import -
Requirements and caveats from upstream:
- Tested on: VPS / 8 GB RAM / 4 vCPU / k3s single-node.
- ### 1. Prerequisites
- **Manual option**: k3s, Docker, and kubectl installed on the VPS
Basic usage or getting-started notes:
- A VPS (or machine) with 4 vCPU / 8 GB RAM minimum
- API keys for your desired AI worker providers
- **Ansible option**: ansible installed locally + SSH root access to the VPS
- Source: https://github.com/hey-intent/patchwork-agents
- Extracted from upstream docs: https://raw.githubusercontent.com/hey-intent/patchwork-agents/HEAD/README.md
## Documentation
- https://github.com/hey-intent/patchwork-agents
## Source
- [Agent Skill Exchange](https://agentskillexchange.com/skills/dispatch-coding-agents-from-github-issue-labels-and-return-pull-requests-with-patchwork-agents/)
Attribution
Comments
Loading comments…
Newsletter
Get the weekly list of safe new skills, and the ones we pulled.