Harden repository for agent-guided contribution workflow

[senolcolak]

Context

The repository structure and AGENTS.md are now in place. The next step is to make CHV consistently usable by human contributors and AI coding agents such as Codex, Claude Code, or similar tools.

CHV is infrastructure-facing software, so the repo needs clear contribution rules around provider boundaries, control-plane behavior, agent behavior, safety checks, and destructive operations. The goal is to keep the workflow practical and lean while borrowing the useful parts of mature Rust repositories such as Turso.

Scope

• Review AGENTS.md against the current repository layout and make sure all paths, crate names, and commands are accurate.
• Ensure README.md and CONTRIBUTING.md point contributors to AGENTS.md where appropriate.
• Add or update lightweight agent/contributor docs, for example:
  • docs/agent-guides/testing.md
  • docs/agent-guides/code-quality.md
  • docs/agent-guides/pr-workflow.md
  • docs/architecture/overview.md
  • docs/security/safety.md, if not already covered elsewhere
• Document the expected local validation commands:
  • cargo fmt --check
  • cargo clippy --workspace --all-features --all-targets -- --deny warnings
  • cargo test --workspace
• Review CI and make sure the same basic quality gates are enforced.
• Add or update PR / issue templates if missing.

CHV-specific notes

Please make the contribution guidance explicit about the main CHV boundaries:

• core VM/domain model
• API/control-plane behavior
• node/agent behavior
• provider implementations such as KVM, Proxmox, or future providers
• storage/network integration
• CLI behavior
• migration and upgrade behavior

Changes that may create, modify, stop, delete, migrate, or reconfigure infrastructure resources should require explicit tests and human review. Destructive operations should be idempotent, clearly named, and guarded by safe defaults.

Acceptance criteria

• AGENTS.md accurately reflects the current CHV repo layout.
• The main build, lint, format, and test commands are documented and work locally.
• CI enforces the same baseline quality gates.
• README.md / CONTRIBUTING.md link to the agent/contributor workflow.
• At least one short agent guide exists for testing and PR workflow.
• Infrastructure-destructive or provider-sensitive areas are clearly marked for human review.
• The repo makes it clear what an AI agent may change and what requires human review.
• The guidance stays lean and does not introduce unnecessary process overhead.