diegosouzapw GitHub Agentic Workflows Skill
Purpose
Comprehensive guidance for creating, deploying, and securing GitHub Agentic Workflows — AI-powered automations using Copilot coding agent, MCP servers, and custom agents for autonomous task execution.
When to Use
- ✅ Orchestrating Copilot coding agent assignments
- ✅ Building multi-step agentic workflows with stacked PRs
- ✅ Configuring MCP servers for agent tools
- ✅ Implementing safe output handling for AI-generated content
- ✅ Securing agentic pipelines against prompt injection
Core Concepts
Copilot Coding Agent Assignment
// Basic assignment
assign_copilot_to_issue({ owner: "Hack23", repo: "European-Parliament-MCP-Server", issue_number: 100 })
// Advanced with base_ref and custom instructions
assign_copilot_to_issue({
owner: "Hack23", repo: "European-Parliament-MCP-Server",
issue_number: 100,
base_ref: "feature/new-tools",
custom_instructions: "Use TypeScript strict mode. Follow MCP protocol. Add Vitest tests with 80%+ coverage."
})
// Direct PR creation with custom agent
create_pull_request_with_copilot({
owner: "Hack23", repo: "European-Parliament-MCP-Server",
title: "Add new MCP tool", body: "Implementation details",
base_ref: "main", custom_agent: "frontend-specialist"
})
// Track progress
get_copilot_job_status({ owner: "Hack23", repo: "European-Parliament-MCP-Server", job_id: "abc123" })
Stacked PR Workflow
Step 1: Data models (PR → main)
└── Step 2: API client (PR → Step 1 branch)
└── Step 3: MCP tools (PR → Step 2 branch)
MCP Server Configuration
MCP server configuration is defined in .github/copilot-mcp.json. Secret references (${{ secrets.* }}) are resolved by the GitHub Copilot runtime — they are not literal JSON values. The MCP client receives actual token values at startup.
Supply chain note: The npx -y @modelcontextprotocol/server-github pattern downloads the latest version on each invocation. For production environments, pin to a specific version (e.g., @modelcontextprotocol/server-github@0.x.y) or vendor the package locally to prevent supply chain attacks when injecting privileged tokens.
See .github/copilot-mcp.json in this repository for the canonical configuration.
OWASP Agentic Security
Threat Mitigation
| Threat | Mitigation |
|---|---|
| Prompt injection | Input validation, output sanitization |
| Excessive agency | Minimal tool permissions, scope limits |
| Data exfiltration | Sandbox environments, network controls |
| Supply chain | Pinned action versions, dependency scanning |
Safe Output Handling
- Validate AI-generated code before merging
- Use CodeQL scanning on Copilot PRs
- Require human review for security-critical changes
- Sanitize all outputs before use in downstream tools