Add AGENTS.md handoff for coding-agent contributors

[EShener]

Hi, Codanna looks like a strong fit for a compact agent handoff file because the project is explicitly built for Claude/Gemini/Codex/Cursor-style assistants and exposes code/document search through MCP.

I ran a local dry run with agent-ready against the repo. It estimated readiness at 75/100 -> 100/100 with the team preset.

The useful, repo-specific pieces look like:

• add a root AGENTS.md that points coding agents to README.md, CONTRIBUTING.md, contributing/README.md, Cargo.toml, CLAUDE.md.example, .mcp_stdio.json, .codannaignore, src/, tests/, and the existing .github/workflows/quick-check.yml / full-test.yml workflows
• document the verification path already described in CONTRIBUTING.md: cargo build --release --all-features, ./contributing/scripts/quick-check.sh, ./contributing/scripts/auto-fix.sh, and ./contributing/scripts/full-test.sh before PRs
• call out safety-sensitive areas for agents: MCP stdio/HTTP transport, document/RAG indexing, embedding model downloads, .codannaignore, install scripts, release packaging, and changes to parser/language support
• add small shims for Claude/Cursor/Gemini/Copilot so different coding-agent tools read the same canonical handoff guidance
• optional readiness workflow can surface missing agent instructions in PRs while reusing the existing CI expectations

This is intentionally not a request to replace your existing static checks; it is just a way to make those existing rules easier for coding agents and human contributors to find before editing.

Reproduce without writing files:

npx --yes @eshen_fox_mie/agent-ready improve --preset team --dry-run --format issue

The check is local-only and does not require API keys. If this direction is useful, I can follow up with a small PR focused only on the handoff docs/readiness workflow. If not, feel free to close.