Copilot Personas

A Drop-in Agentic Framework for GitHub Copilot and VS Code

Transform your repository into an AI-powered development environment with specialized agents, reusable skills, and standardized workflows.

Quick Start

1. Copy .github/ folder to your repository
2. Copy AGENTS.md to your repository root
3. Customize agents and instructions to match your tech stack

What's Inside

.github/agents/ — 8 Specialized AI Personas

Pre-configured agents for every phase of development:

• Product Manager → Requirements & user stories
• Architect → System design & technical decisions
• UI/UX Designer → Visual design & accessibility
• Builder → Implementation & coding
• QA Engineer → Testing & verification
• Reviewer → Code review & quality assurance
• Refactoring Engineer → Code clean-up & optimization
• Site Reliability Engineer → Infrastructure & operations
• Security Auditor → Security compliance & audits

.github/skills/ — Modular Workflows

Reusable capabilities organized by domain:

• architecture/ — System design, API patterns, cloud architecture
• core-engineering/ — Coding standards, testing, debugging
• design/ — UI/UX workflows, accessibility, visual design
• languages/ — Language-specific best practices
• operations/ — DevOps, observability, incident response
• product/ — PRDs, roadmaps, user stories
• security/ — Threat modeling, compliance, auditing

.github/instructions/ — Context-Aware Guidance

Technical standards that auto-activate based on file type and context:

• Tech Strategy → Golden Path for technology choices
• Language Guides → Go, Rust, Python, TypeScript standards
• Architecture Patterns → API design, event-driven systems
• Security Policies → Authentication, secrets management

templates/ — Extensibility

Create new resources using standardized templates:

• agent.md — Define new AI personas
• skill.md — Add reusable workflows
• instruction.md — Create context-aware rules

Key Concepts

Agents & Handoffs

Agents collaborate through handoffs—structured transitions that pass work between personas with explicit artifacts.

Example Flow:

Product Manager → Architect → Builder → QA Engineer → Reviewer
     (PRD)      →  (Design) →  (Code)  →  (Tests)  → (Approval)

Each handoff includes:

• Artifact path (e.g., ./artifacts/prd_feature.md)
• Clear prompt for the next agent
• Verification step to ensure artifact quality

Artifacts & Naming

All planning documents use standardized naming in ./artifacts/:

• prd_[feature].md — Product requirements
• adr_[topic].md — Architecture decisions
• design_spec_[component].md — Visual designs
• roadmap_[project].md — Execution roadmaps
• plan_[task].md — Implementation plans

Skill Activation

Skills auto-load based on patterns defined in .github/skills/skill-rules.json. Agents check this file at initialization to discover relevant capabilities.

Adding New Resources

New Agent

1. Copy templates/agent.md
2. Define identity, responsibilities, and handoffs
3. Register in AGENTS.md

New Skill

1. Copy templates/skill.md
2. Define workflow and patterns
3. Add activation rules to skill-rules.json

New Instruction

1. Copy templates/instruction.md
2. Define standards and compliance requirements
3. Reference from relevant agents

Learn More

• AGENTS.md — Complete agent registry and capability matrix
• .github/copilot-instructions.md — Global AI governance
• Tech Strategy — Golden Path for technology choices