Epic: Automation Layer

[prosdev]

Description

This epic focuses on building automation capabilities on top of the Subagent Foundation. While Epic #2 established read-only context provision agents, this epic will add write operations for automated code changes, PR management, and issue handling.

Philosophy: Automation that's informed by rich context, reducing the risk of AI-generated mistakes.

Goals

1. Implement PR Subagent for automated PR creation and management
2. Add code modification capabilities with safety checks
3. Build issue management automation (labeling, triage, assignment)
4. Create automated testing and validation workflows
5. Implement rollback and safety mechanisms

Acceptance Criteria

• PR subagent can create, update, and merge pull requests via gh CLI
• Code modification agent can safely apply changes with review
• Issue management automation handles common workflows
• All automation includes safety checks and dry-run modes
• Comprehensive test coverage (>80%)
• Documentation includes best practices and safety guidelines

Definition of Done

• All code passes linting, formatting, and type checking
• All tests passing with >= 80% coverage
• Safety mechanisms prevent destructive operations
• Documentation is complete with examples and warnings
• Real-world testing shows automation reduces manual work
• Rollback mechanisms tested and documented

Dependencies

• ✅ Epic Epic: Core Context Provider #1: Core Context Provider (completed)
• ✅ Epic Epic: Subagent Foundation #2: Subagent Foundation (completed)
• Related: Issue Implement Context API and GitHub Integration #5 (Implement Context API and GitHub Integration)

Risks & Mitigations

Risk: Automation could make unintended code changes
Mitigation: Dry-run mode, approval workflows, rollback capability

Risk: GitHub API rate limits
Mitigation: Caching, throttling, incremental operations

Future Enhancements

• Multi-step workflows (analyze → plan → implement → test → PR)
• Learning from past automation outcomes
• Integration with CI/CD pipelines
• Conflict resolution strategies

Priority: Medium (foundation must be solid first)

Status: 📋 Planning - Not yet started

[prosdev]

Status Update:

This epic is on hold. Our current philosophy is to be a context provider, not an automation layer.

What we've built instead:

• dev_plan provides rich context for LLMs to create their own plans
• dev_refs shows code relationships for impact analysis
• dev_map gives codebase overview

The automation (PRs, code changes) is handled by the AI tools (Cursor, Claude Code) using our context. This avoids duplicating their capabilities.

Moving to "Shelved" - may revisit if there's demand for standalone automation.