End-to-End PR Lifecycle Manager for AI Coding Agents

You are an expert in AI-powered software development workflows. Design a complete end-to-end PR lifecycle management system where coding agents handle everything from idea to merge with minimal human intervention. ## System Design ### 1. Task Intake & Planning - Parse natural language task descriptions or GitHub issues - Break down into atomic, parallelizable sub-tasks - Estimate complexity and assign to appropriate agent - Generate a task dependency graph (Mermaid diagram) ### 2. Code Generation Phase - Agent workspace isolation strategy (branch naming, worktree vs clone) - Context gathering: which files to read, how to build codebase understanding - Implementation with test-first approach - Self-review checklist before PR creation ### 3. Automated Review - Static analysis integration (linting, type checking, security scan) - AI reviewer prompt: logic errors, performance issues, style violations - Review comment format: severity level + suggestion + code snippet - Auto-fix for trivial issues ### 4. Conflict Resolution - Semantic conflict detection (not just git merge conflicts) - Resolution priority rules for multi-agent scenarios - Human escalation criteria ### 5. Merge Strategy - Quality gates: tests, coverage threshold, review approval - Merge queue management - Rollback automation if post-merge CI fails ### 6. Observability - Metrics: time-to-merge, agent success rate, human intervention rate - Dashboard design Please generate: - Complete workflow config - Agent system prompts for each phase - GitHub Actions workflow - Mermaid diagram of the full lifecycle Repo context: [DESCRIBE YOUR REPO] Team workflow: [CURRENT PR PROCESS]