# Azhou Verify

> Run and report the authoritative Azhou AI Hub repository verification gate. Use before completion, handoff, commit, pull request, release, or when full-codebase evidence is requested.

- Skill: `tefuirnever/azhou-verify` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add tefuirnever/azhou-verify`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tefuirnever/azhou-verify/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: TeFuirnever (https://skillmd.com/u/tefuirnever)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/tefuirnever/azhou-verify

---


# Azhou Verify

**🦊 阿舟 · Azhou Verify**

> 🧪 完整 gate 跑完，结论才成立。

Use the repository Foundation CLI as the only full-gate entry. Do not replace it with selected tests or infer success from prior output.

## Brand protocol

Emit this exact display event once, at the start of every run, with the resolved checkout scope:

```text
🦊 阿舟 · Azhou Verify 启动｜mode=verify｜scope=<checkout>
```

Use `✅ 验证通过` only after the full gate exits successfully and its output is read back. Use `❌ 验证失败` for any failed or unavailable gate and `🔒 阿舟暂停这一项` when an explicit checkout is missing. Emoji is display-only; keep JSON keys, schema values, digests, paths, commands, test names, and raw evidence emoji-free. A host without Unicode may remove the leading emoji while preserving the fixed text, `｜` separators, fields, and values.

## Workflow

1. Resolve the checkout from a user-supplied path, or from the current Git root only when both `scripts/azhou_hub.py` and `docs/skill-standard.md` exist. Do not scan unrelated directories or infer a harness home.
2. Run `python scripts/azhou_hub.py verify` from that checkout. This is the public, reproducible integrity gate and still recomputes the approved Super Caveman exact diff against the current staged or committed tree. Use `--python <interpreter>` only when the user or environment requires an explicit interpreter.
3. Add `--promotion-evidence` only for an explicitly requested maintainer/release replay after both required Git-external Super Caveman records are available. This mode authenticates those raw records against the same exact diff. Never describe the default public gate as authenticated human-promotion evidence.
4. Derive the verdict mechanically: the gate is **green** only when the exit code is 0 and the captured output ends with the gate's own passing summary; a failure verdict line anywhere in the captured output makes it **red** regardless of any other wording. Quote the exit code and the last captured line in the receipt `verification` field. A green gate takes the success anchor; a red or unavailable gate takes the failure anchor with the conflicting lines quoted. Skipped and unavailable checks appear in the report as skipped and unavailable.
5. A failed gate blocks a completion claim but does not authorize fixes outside the user's task.
6. End with a receipt containing `schema`, `status`, `mode`, `scope`, `command`, `changes`, `verification`, `holds`, and `next_action`. `changes` is always empty.

For requirements and the exact full-gate boundary, read [setup and compatibility](references/setup.md).