universal-dev-standards 6.13.0-beta.4 → 6.13.0

package/bundled/ai/standards/turn-completion-integrity.ai.yaml

@@ -3,8 +3,8 @@
 
 id: turn-completion-integrity
 meta:
-  version: "1.4.0"
-  updated: "2026-09-26"
+  version: "1.4.1"
+  updated: "2026-09-28"
   source: core/turn-completion-integrity.md
   description: An agent must not end a turn having stated a next action it did not take; enforced at turn end, not by instruction
   related:
@@ -134,12 +134,18 @@ supported_harnesses:
     config_file: ".codex/hooks.json"
     block_shape: '{"decision":"block","reason":...}, exit 0 — plain text or empty stdout documented as invalid for this event'
     script: scripts/hooks/check-turn-completion-codex.mjs
+    activation: "Codex skips a project hook until the project is trusted AND the exact hook definition is trusted via /hooks (trust keyed to the definition's hash). Under codex exec an untrusted hook is skipped silently. Measured 2026-09-28, codex-cli 0.155.1"
     known_limit: "R9 (human-directed stop exemption) is best-effort — Codex gives the assistant's last message directly but not the human's; a failed transcript read leaves the human side empty rather than skipping detection"
   - harness: gemini-cli
     event: AfterAgent
     config_file: ".gemini/settings.json"
     block_shape: '{"decision":"deny","reason":...}, exit 0 — the documented preferred path over exit code 2'
     script: scripts/hooks/check-turn-completion-gemini.mjs
+    status: legacy
+    reason: "Google retired Gemini CLI for personal accounts on 2026-06-18 in favour of Antigravity CLI; enterprise accounts keep both. Adapter kept for them; never confirmed against a real Gemini CLI session"
+  - harness: antigravity-cli
+    status: not-yet-supported
+    reason: "Documented Stop contract differs: config .agents/hooks.json, payload has only transcriptPath (no final or human message), block is {\"decision\":\"continue\",\"reason\":...}. Adapter waits until that contract is observed against a real session (R3)"
   - harness: cursor
     status: evaluated-not-supported
     reason: "Whether Cursor's stop hook can actually block a turn was unresolved as of writing; shipping an adapter against an unverified contract repeats the exact failure R3 exists to prevent"

package/bundled/core/turn-completion-integrity.md

@@ -2,8 +2,8 @@
 
 > **Language**: English | [繁體中文](../locales/zh-TW/core/turn-completion-integrity.md)
 
-**Version**: 1.4.0
-**Last Updated**: 2026-09-25
+**Version**: 1.4.1
+**Last Updated**: 2026-09-28
 **Applicability**: Any harness where an agent ends a turn and hands control back to a human
 **Scope**: universal
 **Industry Standards**: none claimed — derived from observed failures, see Evidence
@@ -149,13 +149,22 @@ prevent, one level up.
 ## Supported harnesses
 
 The check is enforced only where a harness adapter exists and a hook is
-actually wired into that harness's own config. As of v1.4.0:
+actually wired into that harness's own config. As of v1.4.1:
 
 | Harness | Event | Config file | Block contract |
 |---|---|---|---|
 | Claude Code | Stop | `.claude/settings.json` | stdout `{"decision":"block","reason":...}`, exit 0; silence allows |
 | Codex | Stop | `.codex/hooks.json` | stdout `{"decision":"block","reason":...}`, exit 0 — plain text or empty stdout is documented as invalid for this event |
-| Gemini CLI | AfterAgent | `.gemini/settings.json` | stdout `{"decision":"deny","reason":...}`, exit 0 — the documented preferred path over exit code 2 |
+| Gemini CLI (legacy) | AfterAgent | `.gemini/settings.json` | stdout `{"decision":"deny","reason":...}`, exit 0 — the documented preferred path over exit code 2 |
+
+Wired is not running on Codex. Codex skips a project hook until the project
+is trusted **and** that exact hook definition has been trusted through `/hooks`
+in an interactive Codex session; trust is recorded against the definition's
+hash, so a changed definition needs trusting again. Under `codex exec` an
+untrusted hook is skipped with no message at all (measured 2026-09-28,
+codex-cli 0.155.1: the adapter was never invoked until the hook was trusted,
+then blocked and allowed exactly as its tests say). `uds init --with-hooks`
+prints this next to the install line.
 
 Codex's R9 exemption is best-effort, not silent failure: Codex's Stop payload
 gives the assistant's final message directly but not the human's, so reading
@@ -172,6 +181,18 @@ shape) still leaves the human side empty rather than throwing — detection
 still runs on the assistant's message, only the R9 exemption for that one
 turn may be missed.
 
+Gemini CLI is legacy. Google retired it for personal accounts on
+2026-06-18 in favour of Antigravity CLI (`agy`); enterprise accounts keep
+access to both. The adapter stays for those users, but it has never been
+confirmed against a real Gemini CLI session, and new adopters on Google's
+tooling should expect Antigravity CLI, which is **not yet supported**. Its
+documented Stop hook contract differs from every adapter above in the ways
+that matter: the hook is configured in `.agents/hooks.json`, the payload
+carries only a `transcriptPath` (no final message, no human message), and a
+block is `{"decision":"continue","reason":...}`, not `block` or `deny`. An
+adapter will be added once that contract has been observed against a real
+session — the same reason Cursor below has none.
+
 Cursor was evaluated and is not supported: whether its stop hook can actually
 block a turn in the way this standard requires was unresolved as of this
 writing, and shipping an adapter against an unverified contract would repeat

package/bundled/hooks/check-turn-completion.mjs

@@ -6,7 +6,8 @@
  * commitment to a next action that the turn then ended without taking.
  *
  * Contract (Claude Code Stop hook):
- *   stdin  — JSON with session_id, transcript_path, stop_hook_active
+ *   stdin  — JSON with session_id, transcript_path, stop_hook_active, and
+ *            (current Claude Code) last_assistant_message
  *   block  — print {"decision":"block","reason":"..."} on stdout, exit 0
  *   allow  — print nothing, exit 0
  *
@@ -21,6 +22,17 @@
  * is reading Claude Code's stdin/transcript shape and writing Claude Code's
  * output shape.
  *
+ * The agent's final message is taken from stdin's `last_assistant_message`
+ * when present, NOT from the transcript. Measured 2026-09-28 against Claude
+ * Code 2.1.283 in a live `claude -p` session: at the moment the Stop hook runs,
+ * the transcript does not yet contain the final assistant message — the
+ * adapter read "" and allowed 5 of 5 turns that should have blocked. (In a
+ * longer session it would have read the PREVIOUS turn's message instead.)
+ * Every self-test and unit test passed throughout, because they all hand the
+ * adapter a transcript that is already complete. The transcript is still read
+ * for the human's side (R9), which is written at the start of the turn, and
+ * as the fallback for Claude Code versions that do not send the field.
+ *
  * Usage: node check-turn-completion.mjs  (reads stdin)
  *        node check-turn-completion.mjs --self-test
  *        node check-turn-completion.mjs --languages
@@ -82,15 +94,19 @@ async function main() {
   if (!data || typeof data !== 'object') return;
   if (data.stop_hook_active === true) return;
 
-  const tp = data.transcript_path;
-  if (!tp || !existsSync(tp)) return;   // cannot tell is not the same as should block
+  const fromStdin = typeof data.last_assistant_message === 'string' ? data.last_assistant_message : null;
 
-  let msgs;
-  try { msgs = lastMessages(tp); } catch { return; }
+  const tp = data.transcript_path;
+  let msgs = { assistant: '', user: '' };
+  if (tp && existsSync(tp)) {
+    try { msgs = lastMessages(tp); } catch { /* human side unknown; see below */ }
+  } else if (fromStdin === null) {
+    return;   // no message from either source: cannot tell is not the same as should block
+  }
 
   const verdict = await decide({
     sessionId: data.session_id,
-    assistantText: msgs.assistant,
+    assistantText: fromStdin !== null ? fromStdin : msgs.assistant,
     userText: msgs.user,
   });
   if (!verdict.fire) return;

package/bundled/locales/zh-CN/CHANGELOG.md

@@ -1,7 +1,7 @@
 ---
 source: ../../CHANGELOG.md
-source_version: 6.13.0-beta.4
-translation_version: 6.13.0-beta.4
+source_version: 6.13.0
+translation_version: 6.13.0
 last_synced: 2026-09-27
 status: current
 ---
@@ -17,6 +17,23 @@ status: current
 
 ## [Unreleased]
 
+## [6.13.0] - 2026-09-28
+
+> **正式版**：包含下方 6.13.0-beta.1 至 beta.5 的全部内容，beta.5 之后没有任何变更。重点：回合收尾关卡（agent 说了下一步却没做就不得结束回合）现在覆盖 **Claude Code 与 Codex**，两者都以从 npm 安装的版本在真实会话中验证过（Codex 要先用 `/hooks` 信任才会执行）；Claude Code 关卡真的会拦（到 beta.4 为止每个回合都放行）；Windows 上可用；`uds uninstall` 会移除它；`uds init` 让提交前检查真的执行；`developer-memory` 1.2.0。Gemini CLI 标为过时（Google 已对个人账号停用，改由 Antigravity CLI 取代，后者尚未支持）。
+
+## [6.13.0-beta.5] - 2026-09-28
+
+> **测试版** — 以 `npm install -g universal-dev-standards@beta` 安装。要测什么、已知限制、如何退回正式版：见 [docs/PRE-RELEASE.md](../../docs/PRE-RELEASE.md)。**装过 6.13.0-beta.1～beta.4 的请升级：那几版的 Claude Code 回合收尾关卡从未拦下过。**
+
+### 修复
+
+- **Claude Code 的回合收尾关卡在真实的 Claude Code 会话里从未拦下过任何回合——6.13.0-beta.1 到 beta.4 的每个采用者都接上了、也在执行，而它放行了每一个回合。** 2026-09-28 以 Claude Code 2.1.283 在真实 `claude -p` 会话实测：Stop hook 执行的那一刻，对话记录里还没有最后一条 AI 回复，因此只读对话记录的适配层看到空消息，应拦的 5 次全部放行；在较长的对话里，它判断的会是上一轮的回复。自我测试与单元测试从头到尾都通过，因为每一个都喂给适配层一份已经写完的对话记录。适配层现在改从 stdin 的 `last_assistant_message`（Claude Code 会发送）取得最后一条回复，对话记录只用来读人的那一侧（R9），以及在不发送这个字段的版本上作为备援。修复后已再次实测。新测试重现关卡当下实际看到的形状，在旧适配层上会失败。
+- **Codex 的关卡装上了却从未执行，而且没有任何提示。** Codex 会跳过项目级的 hook，直到项目被信任、且这一支 hook 通过 `/hooks` 被信任为止——在 `codex exec` 下完全无声。`uds init --with-hooks` 现在会在 Codex 安装那一行旁边说明这件事，标准的〈支持的执行环境〉与 `docs/PRE-RELEASE.md` 也补上这个步骤。信任之后，Codex 适配层的拦截与放行与它的测试完全一致（2026-09-28 实测，codex-cli 0.155.1）。
+
+### 变更
+
+- **`turn-completion-integrity` 1.4.1：Gemini CLI 适配层标为过时，并明列 Antigravity CLI 尚未支持。** Google 于 2026-06-18 对个人账号停用 Gemini CLI，改由 Antigravity CLI（`agy`）取代；企业账号两者都还能用。适配层为他们保留，但从未在真实的 Gemini CLI 会话中验证过，因此不再与 Claude Code、Codex 适配层并列。Antigravity CLI 文档记载的 Stop hook 契约与所有已发布的适配层都不同（配置在 `.agents/hooks.json`；传入数据只有 `transcriptPath`，没有最后一条回复或人的消息；拦截是 `{"decision":"continue"}`），因此在真实会话中观察到这份契约之前不发布适配层——与 Cursor 不支持是同一条规则。行为不变：选了 Gemini CLI 时，`uds init --with-hooks` 仍会接上它的关卡。
+
 ## [6.13.0-beta.4] - 2026-09-28
 
 > **测试版** — 以 `npm install -g universal-dev-standards@beta` 安装。要测什么、已知限制、如何退回正式版：见 [docs/PRE-RELEASE.md](../../docs/PRE-RELEASE.md)。注意：6.13.0-beta.3 从未上架 npm，其内容随本版发布。

package/bundled/locales/zh-CN/README.md

@@ -15,7 +15,7 @@ status: current
 
 > **语言**: [English](../../README.md) | [繁體中文](../zh-TW/README.md) | 简体中文
 
-**版本**: 6.13.0-beta.4 (Pre-release) | **发布日期**: 2026-09-28 | **授权**: [双重授权](../../LICENSE) (CC BY 4.0 + MIT)
+**版本**: 6.13.0 | **发布日期**: 2026-09-28 | **授权**: [双重授权](../../LICENSE) (CC BY 4.0 + MIT)
 
 语言无关、框架无关的软件项目文档标准。通过 AI 原生工作流，确保不同技术栈之间的一致性、质量和可维护性。
 

package/bundled/locales/zh-CN/SECURITY.md

@@ -13,8 +13,7 @@ status: current
 <!-- UDS_SUPPORTED_VERSIONS_START -->
 | 版本 | 支持状态 |
 |------|--------|
-| 6.13.0-beta.4 | ✅ 预发布版本 |
-| 6.12.0 | ✅ 最新正式版 |
+| 6.13.0 | ✅ 最新正式版 |
 | < 6.0.0 | ❌ 已终止支持 |
 <!-- UDS_SUPPORTED_VERSIONS_END -->
 

package/bundled/locales/zh-CN/core/turn-completion-integrity.md

@@ -1,9 +1,9 @@
 ---
 source: ../../../core/turn-completion-integrity.md
-source_version: 1.4.0
-translation_version: 1.4.0
-last_synced: 2026-09-26
-source_hash: 401b74843abc
+source_version: 1.4.1
+translation_version: 1.4.1
+last_synced: 2026-09-28
+source_hash: 0c04676006c0
 status: current
 ---
 
@@ -11,8 +11,8 @@ status: current
 
 > **语言**: [English](../../../core/turn-completion-integrity.md) | [繁體中文](../../zh-TW/core/turn-completion-integrity.md) | 简体中文
 
-**版本**: 1.4.0
-**最后更新**: 2026-09-25
+**版本**: 1.4.1
+**最后更新**: 2026-09-28
 **适用范围**: 任何由 agent 结束回合、把控制权交还给人的执行环境
 **Scope**: universal
 **行业标准**: 不声称任何来源——由实际观察到的失败归纳，见「证据」
@@ -144,13 +144,19 @@ agent 写下「我接着做 X」，然后结束回合，而 X 没有做。
 ## 支持的执行环境
 
 这个检查只在「适配层存在，且 hook 真的被接入该执行环境自己的配置」时才生效。
-截至 v1.4.0：
+截至 v1.4.1：
 
 | 执行环境 | 事件 | 配置文件 | 拦截契约 |
 |---|---|---|---|
 | Claude Code | Stop | `.claude/settings.json` | stdout 输出 `{"decision":"block","reason":...}`，exit 0；沉默即放行 |
 | Codex | Stop | `.codex/hooks.json` | stdout 输出 `{"decision":"block","reason":...}`，exit 0——官方文档写明这个事件纯文本或空输出无效 |
-| Gemini CLI | AfterAgent | `.gemini/settings.json` | stdout 输出 `{"decision":"deny","reason":...}`，exit 0——官方文档标记为优先于 exit code 2 的做法 |
+| Gemini CLI（过时） | AfterAgent | `.gemini/settings.json` | stdout 输出 `{"decision":"deny","reason":...}`，exit 0——官方文档标记为优先于 exit code 2 的做法 |
+
+在 Codex 上，接上了不等于会执行。Codex 会跳过项目级的 hook，直到项目被信任、**而且**
+这一支 hook 的定义在交互式 Codex 会话里通过 `/hooks` 被信任为止；信任记录绑定在定义的
+哈希值上，定义一改就要重新信任。在 `codex exec` 下，未被信任的 hook 会被跳过，而且完全
+没有任何消息（2026-09-28 实测，codex-cli 0.155.1：hook 被信任之前适配层一次都没被调用；
+信任之后，拦截与放行都与它的测试完全一致）。`uds init --with-hooks` 会在安装那一行旁边打印这件事。
 
 Codex 的 R9 豁免是尽力而为，不是静默失效：Codex 的 Stop payload 直接给出
 agent 的最后一条消息，却不给出用户的；要拿到用户那一侧必须解析一份
@@ -165,6 +171,15 @@ agent 的最后一条消息，却不给出用户的；要拿到用户那一侧
 用户那一侧变空、不会抛出异常——检测仍照样运行在 agent 消息上，只有那一轮
 的 R9 豁免可能漏掉。
 
+Gemini CLI 已过时。Google 于 2026-06-18 对个人账号停用 Gemini CLI，
+改由 Antigravity CLI（`agy`）取代；企业账号两者都还能用。这个适配层为那些用户保留，
+但它从未在真实的 Gemini CLI 会话中验证过；使用 Google 工具的新采用者应预期的是
+Antigravity CLI，而它**尚未支持**。它文档记载的 Stop hook 契约，在关键之处与上表每一个
+适配层都不同：hook 配置在 `.agents/hooks.json`、传入数据只有 `transcriptPath`
+（没有最后一条回复、也没有人的消息）、拦截是 `{"decision":"continue","reason":...}`
+而不是 `block` 或 `deny`。等这份契约在真实会话中观察到之后才会加入适配层——
+与下方 Cursor 没有适配层是同一个理由。
+
 Cursor 已评估但不支持：截至撰写本文时，Cursor 的 stop hook 能不能真的
 拦下一个回合仍未确定，若对着一个没人验证过的契约交付一份适配层，
 等于重演 R3 要防的那个失败——一个没人确认过真的在执法的执法机制。

package/bundled/locales/zh-TW/CHANGELOG.md

@@ -1,7 +1,7 @@
 ---
 source: ../../CHANGELOG.md
-source_version: 6.13.0-beta.4
-translation_version: 6.13.0-beta.4
+source_version: 6.13.0
+translation_version: 6.13.0
 last_synced: 2026-09-27
 status: current
 ---
@@ -17,6 +17,23 @@ status: current
 
 ## [Unreleased]
 
+## [6.13.0] - 2026-09-28
+
+> **正式版**：包含下方 6.13.0-beta.1 至 beta.5 的全部內容，beta.5 之後沒有任何變更。重點：回合收尾關卡（agent 說了下一步卻沒做就不得結束回合）現在涵蓋 **Claude Code 與 Codex**，兩者都以從 npm 安裝的版本在真實工作階段驗證過（Codex 要先用 `/hooks` 信任才會執行）；Claude Code 關卡真的會擋（到 beta.4 為止每個回合都放行）；Windows 上可用；`uds uninstall` 會移除它；`uds init` 讓提交前檢查真的執行；`developer-memory` 1.2.0。Gemini CLI 標為過時（Google 已對個人帳號停用，改由 Antigravity CLI 取代，後者尚未支援）。
+
+## [6.13.0-beta.5] - 2026-09-28
+
+> **測試版** — 以 `npm install -g universal-dev-standards@beta` 安裝。要測什麼、已知限制、如何退回正式版：見 [docs/PRE-RELEASE.md](../../docs/PRE-RELEASE.md)。**裝過 6.13.0-beta.1～beta.4 的請升級：那幾版的 Claude Code 回合收尾關卡從未擋下過。**
+
+### 修正
+
+- **Claude Code 的回合收尾關卡在真實的 Claude Code 工作階段裡從未擋下過任何回合——6.13.0-beta.1 到 beta.4 的每個採用者都接上了、也在執行，而它放行了每一個回合。** 2026-09-28 以 Claude Code 2.1.283 在真實 `claude -p` 工作階段實測：Stop hook 執行的那一刻，逐字稿裡還沒有最後一則 AI 回覆，因此只讀逐字稿的適配層看到空訊息，應擋的 5 次全部放行；在較長的對話裡，它判斷的會是上一輪的回覆。自我測試與單元測試從頭到尾都通過，因為每一支都餵給適配層一份已經寫完的逐字稿。適配層現在改從 stdin 的 `last_assistant_message`（Claude Code 會送）取得最後一則回覆，逐字稿只用來讀人的那一側（R9），以及在不送這個欄位的版本上當備援。修正後已再次實測。新測試重現關卡當下實際看到的形狀，在舊適配層上會失敗。
+- **Codex 的關卡裝上了卻從未執行，而且沒有任何提示。** Codex 會略過專案層級的 hook，直到專案被信任、且這一支 hook 透過 `/hooks` 被信任為止——在 `codex exec` 底下完全無聲。`uds init --with-hooks` 現在會在 Codex 安裝那一行旁邊說明這件事，標準的〈支援的執行環境〉與 `docs/PRE-RELEASE.md` 也補上這個步驟。信任之後，Codex 適配層的擋與放行與它的測試完全一致（2026-09-28 實測，codex-cli 0.155.1）。
+
+### 變更
+
+- **`turn-completion-integrity` 1.4.1：Gemini CLI 適配層標為過時，並明列 Antigravity CLI 尚未支援。** Google 於 2026-06-18 對個人帳號停用 Gemini CLI，改由 Antigravity CLI（`agy`）取代；企業帳號兩者都還能用。適配層為他們保留，但從未在真實的 Gemini CLI 工作階段中驗證過，因此不再與 Claude Code、Codex 適配層並列。Antigravity CLI 文件記載的 Stop hook 契約與所有已出貨的適配層都不同（設定在 `.agents/hooks.json`；傳入資料只有 `transcriptPath`，沒有最後一則回覆或人的訊息；攔截是 `{"decision":"continue"}`），因此在真實工作階段中觀察到這份契約之前不出貨適配層——與 Cursor 不支援是同一條規則。行為不變：選了 Gemini CLI 時，`uds init --with-hooks` 仍會接上它的關卡。
+
 ## [6.13.0-beta.4] - 2026-09-28
 
 > **測試版** — 以 `npm install -g universal-dev-standards@beta` 安裝。要測什麼、已知限制、如何退回正式版：見 [docs/PRE-RELEASE.md](../../docs/PRE-RELEASE.md)。注意：6.13.0-beta.3 從未上架 npm，其內容隨本版出貨。

package/bundled/locales/zh-TW/README.md

@@ -15,7 +15,7 @@ status: current
 
 > **語言**: [English](../../README.md) | 繁體中文 | [简体中文](../zh-CN/README.md)
 
-**版本**: 6.13.0-beta.4 (Pre-release) | **發布日期**: 2026-09-28 | **授權**: [雙重授權](../../LICENSE) (CC BY 4.0 + MIT)
+**版本**: 6.13.0 | **發布日期**: 2026-09-28 | **授權**: [雙重授權](../../LICENSE) (CC BY 4.0 + MIT)
 
 語言無關、框架無關的軟體專案文件標準。透過 AI 原生工作流，確保不同技術堆疊之間的一致性、品質和可維護性。
 

package/bundled/locales/zh-TW/SECURITY.md

@@ -13,8 +13,7 @@ status: current
 <!-- UDS_SUPPORTED_VERSIONS_START -->
 | 版本 | 支援狀態 |
 |------|--------|
-| 6.13.0-beta.4 | ✅ 預發布版本 |
-| 6.12.0 | ✅ 最新正式版 |
+| 6.13.0 | ✅ 最新正式版 |
 | < 6.0.0 | ❌ 已終止支援 |
 <!-- UDS_SUPPORTED_VERSIONS_END -->
 

package/bundled/locales/zh-TW/core/turn-completion-integrity.md

@@ -1,9 +1,9 @@
 ---
 source: ../../../core/turn-completion-integrity.md
-source_version: 1.4.0
-translation_version: 1.4.0
-last_synced: 2026-09-26
-source_hash: 401b74843abc
+source_version: 1.4.1
+translation_version: 1.4.1
+last_synced: 2026-09-28
+source_hash: 0c04676006c0
 status: current
 ---
 
@@ -11,8 +11,8 @@ status: current
 
 > **Language**: [English](../../../core/turn-completion-integrity.md) | 繁體中文
 
-**版本**: 1.4.0
-**最後更新**: 2026-09-25
+**版本**: 1.4.1
+**最後更新**: 2026-09-28
 **適用範圍**: 任何由 agent 結束回合、把控制權交還給人的執行環境
 **Scope**: universal
 **產業標準**: 不宣稱任何來源——由實際觀察到的失敗歸納，見「證據」
@@ -144,13 +144,19 @@ agent 寫下「我接著做 X」，然後結束回合，而 X 沒有做。
 ## 支援的執行環境
 
 這個檢查只在「轉接層存在，且 hook 真的被接進該執行環境自己的設定」時才生效。
-截至 v1.4.0：
+截至 v1.4.1：
 
 | 執行環境 | 事件 | 設定檔 | 阻擋契約 |
 |---|---|---|---|
 | Claude Code | Stop | `.claude/settings.json` | stdout 印 `{"decision":"block","reason":...}`，exit 0；沉默即放行 |
 | Codex | Stop | `.codex/hooks.json` | stdout 印 `{"decision":"block","reason":...}`，exit 0——官方文件寫明這個事件純文字或空輸出無效 |
-| Gemini CLI | AfterAgent | `.gemini/settings.json` | stdout 印 `{"decision":"deny","reason":...}`，exit 0——官方文件標記為優先於 exit code 2 的做法 |
+| Gemini CLI（過時） | AfterAgent | `.gemini/settings.json` | stdout 印 `{"decision":"deny","reason":...}`，exit 0——官方文件標記為優先於 exit code 2 的做法 |
+
+在 Codex 上，接上了不等於會執行。Codex 會略過專案層級的 hook，直到專案被信任、**而且**
+這一支 hook 的定義在互動式 Codex 工作階段裡透過 `/hooks` 被信任為止；信任紀錄綁在定義的
+雜湊值上，定義一改就要重新信任。在 `codex exec` 底下，未被信任的 hook 會被略過，而且完全
+沒有任何訊息（2026-09-28 實測，codex-cli 0.155.1：hook 被信任之前適配層一次都沒被呼叫；
+信任之後，擋與放行都與它的測試完全一致）。`uds init --with-hooks` 會在安裝那一行旁邊印出這件事。
 
 Codex 的 R9 豁免是盡力而為，不是靜默失效：Codex 的 Stop payload 直接給
 agent 的最後一則訊息，卻不給使用者的；要拿到使用者那一側必須解析一份
@@ -165,6 +171,15 @@ agent 的最後一則訊息，卻不給使用者的；要拿到使用者那一
 使用者那一側變空、不會拋出例外——偵測仍照樣跑在 agent 訊息上，只有那一輪
 的 R9 豁免可能漏掉。
 
+Gemini CLI 已過時。Google 於 2026-06-18 對個人帳號停用 Gemini CLI，
+改由 Antigravity CLI（`agy`）取代；企業帳號兩者都還能用。這個適配層為那些使用者保留，
+但它從未在真實的 Gemini CLI 工作階段中驗證過；使用 Google 工具的新採用者該預期的是
+Antigravity CLI，而它**尚未支援**。它文件記載的 Stop hook 契約，在關鍵之處與上表每一個
+適配層都不同：hook 設定在 `.agents/hooks.json`、傳入資料只有 `transcriptPath`
+（沒有最後一則回覆、也沒有人的訊息）、攔截是 `{"decision":"continue","reason":...}`
+而不是 `block` 或 `deny`。等這份契約在真實工作階段中觀察到之後才會加入適配層——
+與下方 Cursor 沒有適配層是同一個理由。
+
 Cursor 已評估但不支援：截至撰寫本文時，Cursor 的 stop hook 能不能真的
 擋下一個回合仍未確定，若對著一個沒人驗證過的契約出一份轉接層，
 等於重演 R3 要防的那個失敗——一個沒人確認過真的在執法的執法機制。

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "universal-dev-standards",
-  "version": "6.13.0-beta.4",
+  "version": "6.13.0",
   "description": "CLI tool for adopting Universal Development Standards",
   "keywords": [
     "documentation",

package/src/commands/init.js

@@ -293,6 +293,11 @@ export async function initCommand(options) {
       const codexResult = installCodexHooks(projectPath);
       if (codexResult.installed) {
         console.log(chalk.green('  ✓ Codex Stop hook installed (turn-completion-integrity)'));
+        // Measured 2026-09-28 (codex-cli 0.155.1): Codex skips a project hook
+        // until the project is trusted AND that exact hook definition is
+        // trusted via /hooks — silently under `codex exec`. Installed is not
+        // running; say so here, where the adopter believes it now works.
+        console.log(chalk.yellow('    ⚠ Codex will not run it until you trust it: open Codex in this project, trust the project, then run /hooks and trust this hook. Until then it is skipped (silently under `codex exec`).'));
       } else {
         console.log(chalk.yellow(`  ⚠ Codex hook not installed — ${codexResult.reason}`));
       }
@@ -301,7 +306,7 @@ export async function initCommand(options) {
       const { installGeminiHooks } = await import('../installers/hooks-installer.js');
       const geminiResult = installGeminiHooks(projectPath);
       if (geminiResult.installed) {
-        console.log(chalk.green('  ✓ Gemini CLI AfterAgent hook installed (turn-completion-integrity)'));
+        console.log(chalk.green('  ✓ Gemini CLI AfterAgent hook installed (turn-completion-integrity; Gemini CLI is legacy — see the standard)'));
       } else {
         console.log(chalk.yellow(`  ⚠ Gemini CLI hook not installed — ${geminiResult.reason}`));
       }

package/standards-registry.json

@@ -1,6 +1,6 @@
 {
   "$schema": "https://json-schema.org/draft/2020-12/schema",
-  "version": "6.13.0-beta.4",
+  "version": "6.13.0",
   "lastUpdated": "2026-05-13",
   "description": "Standards registry for universal-dev-standards with integrated skills and AI-optimized formats",
   "formats": {
@@ -58,14 +58,14 @@
     "standards": {
       "name": "universal-dev-standards",
       "url": "https://github.com/AsiaOstrich/universal-dev-standards",
-      "version": "6.13.0-beta.4"
+      "version": "6.13.0"
     },
     "skills": {
       "name": "universal-dev-standards",
       "url": "https://github.com/AsiaOstrich/universal-dev-standards",
       "localPath": "skills",
       "rawUrl": "https://raw.githubusercontent.com/AsiaOstrich/universal-dev-standards/main/skills",
-      "version": "6.13.0-beta.4",
+      "version": "6.13.0",
       "note": "Skills are now included in the main repository under skills/"
     }
   },
@@ -2295,7 +2295,7 @@
       "id": "license-compliance",
       "name": "License Compliance Standards",
       "nameZh": "授權合規標準",
-      "version": "6.13.0-beta.4",
+      "version": "6.13.0",
       "source": {
         "human": "core/license-compliance.md",
         "ai": "ai/standards/license-compliance.ai.yaml"
@@ -2307,7 +2307,7 @@
       "id": "verification-oracle",
       "name": "Verification Oracle Standards",
       "nameZh": "驗證 Oracle 標準",
-      "version": "6.13.0-beta.4",
+      "version": "6.13.0",
       "source": {
         "human": "core/verification-oracle.md",
         "ai": "ai/standards/verification-oracle.ai.yaml"
@@ -2319,7 +2319,7 @@
       "id": "model-provenance",
       "name": "Model Provenance Policy Standards",
       "nameZh": "模型來源政策標準",
-      "version": "6.13.0-beta.4",
+      "version": "6.13.0",
       "source": {
         "human": "core/model-provenance.md",
         "ai": "ai/standards/model-provenance.ai.yaml"
@@ -2331,7 +2331,7 @@
       "id": "resource-cost-boundary",
       "name": "Resource / Cost Boundary Declaration Standards",
       "nameZh": "資源／成本邊界宣告標準",
-      "version": "6.13.0-beta.4",
+      "version": "6.13.0",
       "source": {
         "human": "core/resource-cost-boundary.md",
         "ai": "ai/standards/resource-cost-boundary.ai.yaml"