docs(community): add Xenon Skill and MCP interoperability guide

[xianyu-sheng]

Summary

• add Chinese and English community documentation for reusing VeADK-compatible SKILL.md directories and stdio MCP servers in Xenon
• use the existing examples/codex_with_skill_and_mcp weather example for executable setup and verification commands
• update the reproducible documentation baseline to the published Xenon v0.8.0 release
• document the component mapping and explicitly state that this does not add Agent(runtime="xenon") support

About Xenon v0.8.0

Xenon is an open-source terminal AI coding agent built around an online Evidence Runtime — a deterministic, zero-LLM verification chain that runs vertically through the whole task lifecycle (plan → pre-tool → execution → patch-binding → delivery). Its core contract: an LLM statement is a claim; only a tool result is evidence. This catches "pasted a diff but never wrote the file" hallucinations at delivery time with zero false-kills.

On the SWE-bench_Lite official harness (30 instances × 5 engines, single attempt, cheapest model tier), Xenon reaches 36.7% instance-level with a fully reproducible report (sampling seed, grading method, cost accounting, no-patch attribution) committed to the repo.

Validation

• fumadocs-mdx, next typegen, and tsc --noEmit passed for the original integration change
• the full static site build generated 288/288 pages, including both Xenon routes
• Ark CLI plus Xenon installed and discovered all embedded Skills with 0 errors; the v0.8.0 integration contract, Skill roots, and MCP transports are unchanged
• the repository weather MCP example completed initialize, tools/list, and get_weather
• Xenon v0.8.0 full offline suite: 2088 passed (unit + integration), ruff and compileall green
• v0.8.0 wheel and sdist pass package metadata checks

中文说明

本 PR 增加 Xenon 与 VeADK 在标准 Skill/MCP 组件层的双语互操作指南，并清楚区分「组件复用」和「VeADK 原生 Xenon runtime 适配」。文档以已发布的 Xenon v0.8.0 为稳定版本基线。

Xenon 是一个以「在线 Evidence Runtime 验证链」为核心的开源终端 AI 编程 Agent——LLM 的话只能作为 Claim，工具结果才是 Evidence，验证从事后审计升级为贯穿任务全生命周期的在线拦截。在 SWE-bench_Lite 官方评测（30 实例，最便宜模型档，单次尝试）上达到 36.7% 实例级通过率，完整评测方法与成本已入库可复现。

[xianyu-sheng]

Xenon v0.7.3 baseline update — 2026-07-27

Xenon v0.7.3 is published, and this PR's bilingual guide now targets it in commit 83194f1. The Skill roots, MCP transports, integration CLI schema, and runtime-adapter boundary are unchanged. Current Xenon offline suite: 1662 passed, 38 deselected.