Add AGENTS.md and .github/copilot-instructions.md for AI agent onboarding

[francisfuzz]

Summary

This repository has no agent instruction files, which means every AI coding agent (Copilot, Codex, Cursor, etc.) starts from scratch each time — re-discovering build commands, testing patterns, project structure, and conventions.

Proposal

Add two complementary files:

File	Purpose	Read by
AGENTS.md	Agent-agnostic open standard (agents.md spec)	Copilot cloud agent, OpenAI Codex, Cursor, Amp, Jules, Gemini, and others
.github/copilot-instructions.md	GitHub Copilot-specific instructions (docs)	Copilot Chat, code review, completions, cloud agent

What to include

• Project overview: Go CLI extension for managing stacked PRs
• Build / test / vet commands: go mod download, go build ./..., go test -race -count=1 ./..., go vet ./...
• Project structure: key packages (cmd/, internal/git, internal/stack, internal/github, etc.)
• Testing patterns: config.NewTestConfig(), git.SetOps(mock), MockClient, table-driven tests, testify assertions
• Error handling: typed ExitError codes, errors.As() pattern
• Key interfaces: git.Ops, github.ClientOps
• Coding conventions: Cobra commands, bubbletea TUI, config injection hooks

Design considerations

• copilot-instructions.md should stay under ~4,000 chars (Copilot code review limit)
• AGENTS.md can be more detailed
• Both should use natural language Markdown — no proprietary format
• Content should be broadly applicable (not task-specific)

References

• GitHub docs: Repository custom instructions
• AGENTS.md open standard
• github/awesome-copilot instruction templates