docs: add AI repo intelligence docs under docs/ai

[rovo79]

Motivation

• Provide a compact, machine-friendly set of repository intelligence documents to help contributors and AI agents quickly understand the codebase, runtime entrypoints, stack, architecture, conventions, hotspots, testing, commands, deployment, and security posture.
• Deliver these artifacts under docs/ai/ as requested without changing runtime code so reviewers can use them to plan work or integrate with tooling.

Description

• Added a set of documentation files under docs/ai/ that analyze the repository and point to concrete code locations and assumptions: CODEBASE_MAP.md, STACK.md, ARCHITECTURE.md, CONVENTIONS.md, HOTSPOTS.md, TESTING.md, COMMANDS.md, DEPLOYMENT.md, and SECURITY_AND_RISKS.md.
• Each doc highlights concrete file paths (primarily in packages/server/) and covers entrypoints (packages/server/src/bin/http.ts, packages/server/src/bin/stdio.ts, packages/server/src/index.ts), transport routing (packages/server/src/transports/*), feature modules (packages/server/src/features/*), config and execution helpers (packages/server/src/config.ts, packages/server/src/features/sandboxExecution.ts), and tests (packages/server/src/__tests__/*, packages/server/src/integration/smoke.ts).
• No code or behavior changes were made; these are documentation-only additions and include explicit ‘Assumptions’ sections where the analysis was uncertain or requires external files (CI, secret management, etc.).

Testing

• Ran the package test suite in packages/server using npm test to validate that documentation-only changes did not regress behavior, which runs npm run build and node --test dist/__tests__.
• Result: all tests passed (10 tests / 6 suites, no failures).

---

Codex Task