feat(regress-guard): add AI-coding governance gate plugin

marketplace.json

@@ -740,6 +740,31 @@
         "visualization"
       ],
       "icon": "https://cdn-zcode.z.ai/zcode/official-plugin/assets/obsidian/icon.png"
+    },
+    {
+      "name": "regress-guard",
+      "source": "./plugins/regress-guard",
+      "description": "AI-coding governance gate for ZCode: pre-commit test enforcement, manifest-boundary guard, six-layer trust surface, assumption ledger, and FP-calibrated telemetry.",
+      "description_i18n": {
+        "en": "AI-coding governance gate for ZCode: pre-commit test enforcement, manifest-boundary guard, six-layer trust surface, assumption ledger, and FP-calibrated telemetry.",
+        "zh-CN": "AI 辅助开发治理门禁：提交前测试强制（不信 AI 自报的 done）、清单边界事前拦截、六层信任面、假设账本与误报校准遥测。"
+      },
+      "version": "1.88.0",
+      "author": {
+        "name": "1339190177",
+        "url": "https://github.com/1339190177"
+      },
+      "category": "developer-tools",
+      "keywords": [
+        "governance",
+        "hooks",
+        "pre-commit",
+        "testing",
+        "regression",
+        "guardrails",
+        "spec-driven",
+        "telemetry"
+      ]
     }
   ]
 }

plugins/regress-guard/.pytest_cache/.gitignore

@@ -0,0 +1,2 @@
+# Created by pytest automatically.
+*

plugins/regress-guard/.pytest_cache/CACHEDIR.TAG

@@ -0,0 +1,4 @@
+Signature: 8a477f597d28d172789f06886806bc55
+# This file is a cache directory tag created by pytest.
+# For information about cache directory tags, see:
+#	https://bford.info/cachedir/spec.html

plugins/regress-guard/.pytest_cache/README.md

@@ -0,0 +1,8 @@
+# pytest cache directory #
+
+This directory contains data from the pytest's cache plugin,
+which provides the `--lf` and `--ff` options, as well as the `cache` fixture.
+
+**Do not** commit this to version control.
+
+See [the docs](https://docs.pytest.org/en/stable/how-to/cache.html) for more information.

plugins/regress-guard/.pytest_cache/v/cache/lastfailed

@@ -0,0 +1,3 @@
+{
+  "tests/test_facts.py::test_record_creates_card_with_frontmatter": true
+}

plugins/regress-guard/.zcode-plugin/plugin.json

@@ -0,0 +1,27 @@
+{
+  "name": "regress-guard",
+  "version": "1.88.0",
+  "description": "AI-coding governance gate for ZCode: pre-commit test enforcement, manifest-boundary guard, six-layer trust surface, assumption ledger, and FP-calibrated telemetry.",
+  "author": {
+    "name": "1339190177",
+    "url": "https://github.com/1339190177"
+  },
+  "license": "MIT",
+  "commands": "commands",
+  "hooks": "hooks",
+  "description_i18n": {
+    "en": "AI-coding governance gate for ZCode: pre-commit test enforcement, manifest-boundary guard, six-layer trust surface, assumption ledger, and FP-calibrated telemetry.",
+    "zh-CN": "AI 辅助开发治理门禁：提交前测试强制（不信 AI 自报的 done）、清单边界事前拦截、六层信任面、假设账本与误报校准遥测。"
+  },
+  "keywords": [
+    "governance",
+    "hooks",
+    "pre-commit",
+    "testing",
+    "regression",
+    "guardrails",
+    "spec-driven",
+    "telemetry"
+  ],
+  "category": "developer-tools"
+}

plugins/regress-guard/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 1339190177
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

plugins/regress-guard/README.md

@@ -0,0 +1,112 @@
+# regress-guard
+
+> AI-coding governance gate for ZCode: **don't trust the agent's "done"** — the gate
+> re-runs your tests itself, blocks edits outside the declared change-scope, and
+> keeps human-only-calibrated telemetry on every interception.
+>
+> Design philosophy first: [docs/PHILOSOPHY.md](docs/PHILOSOPHY.md) (sixteen
+> principles and the real incidents behind them). Features age, principles don't.
+> Full workflow: [docs/WORKFLOW.md](docs/WORKFLOW.md) · 中文文档：[README_CN.md](README_CN.md)
+
+## Why
+
+Agents report "done" confidently and cheaply. This plugin makes "done" a
+machine-verified claim, at the only surface that matters in a direct-to-main
+workflow: **the commit**.
+
+- **Completion re-verification (the core)**: a PreToolUse gate intercepts commit
+  commands and runs the project's own test suite itself — agent self-reports are
+  never trusted. EARS-style acceptance rows must be checked off with evidence
+  before `done` is stamped. Same-tree retries hit a content-keyed result cache
+  (0-second gates).
+- **Boundary guard**: edits/writes to files not declared in the active change
+  manifest are blocked before they happen — the plan's file list *is* the boundary.
+- **Assumption ledger**: when reality falsifies a plan's assumption, the
+  was→reality→evidence record is mandatory.
+- **Six-layer trust surface**: project-level notification channels require a
+  human-granted trust table + content pinning; HOME/env-injection redirection is
+  neutralized at the launcher allowlist. Cloned-repo injection surfaces are closed.
+- **FP-calibrated telemetry**: every interception is adjudicated later by the
+  human (useful / false_positive / ignored) — the agent gets no vote on its own
+  accuracy. Public benchmark: 82 cases / 13 families, see [benchmarks/](benchmarks/).
+- **Self-healing**: session start compares the installed copy against the source
+  repo; version drift is announced, not silently tolerated.
+
+## Install
+
+**Marketplace (recommended)**: install `regress-guard` from the ZCode plugin
+marketplace — hooks activate via the host runtime, and the plugin can be
+toggled/upgraded natively.
+
+**Personal-source track**: add marketplace source
+`github:1339190177/regress-guard` in ZCode settings, then install.
+
+**Dev track**: clone this repo, then `bash install.sh` (or tell the AI
+`/regress:install`). Use one track at a time.
+
+Requirements: ZCode client + Python 3.10+.
+
+## Five-minute start
+
+1. In any project, tell the AI: `/regress:init`
+2. Daily flow: state a requirement → the AI produces a plan card → you
+   **approve / edit / cancel** → it works (with boundary enforcement) → on
+   commit, the gate re-runs tests and blocks on failure or unchecked
+   acceptance rows
+3. Project state and archaeology live in `.regress/`
+
+## Side-effects declaration (transparency contract)
+
+- **Execution**: all hooks are local Python 3.10+ processes (no prebuilt
+  binaries, no obfuscated source)
+- **File writes**: `.regress/` inside projects (manifests/history/ledgers) and
+  `~/.zcode/regress-guard-hooks` (installed copy)
+- **Command execution**: exactly two classes — the gate invoking the project's
+  own test runner (pytest/jest/maven/gradle/go, auto-detected), and notification
+  channel commands (**only for human-granted projects**; untrusted configs fall
+  back to machine level, closing the cloned-repo injection surface)
+- **Network**: zero egress by default. Optional interception notifications via
+  WeCom self-built apps — credentials require human trust-grant and the API base
+  is pinned to the official domain. No MCP, no telemetry upload, no third-party
+  analytics
+- **Hook surface**: PreToolUse (commit gate / boundary guard / execution valve),
+  PostToolUse (manifest bridge), UserPromptSubmit (governance context),
+  SessionStart (self-heal), Stop (end-of-turn notices)
+- **Escapes**: `/regress:bypass N` gives a time-boxed amnesty (audited); each
+  feature has an independent env switch
+
+## Health & verification
+
+`bash validate.sh` (expect 5/5) · `python3 -m pytest tests/ -q` (expect green)
+· post-install smoke check runs automatically with `install.sh`
+
+## License
+
+MIT. Third-party: none at runtime (stdlib only).
+
+<!-- generated: reference start · 本区由 scripts/gen_reference.py 生成，勿手改 -->
+
+**实况（由 gen_reference.py 生成，勿手改本区）**
+
+- 命令：16 个 · hook 注册：11 个事件条目 · 测试函数：565 个
+
+| 命令 | 说明 |
+|---|---|
+| `/regress:bypass` | 紧急绕过回归卡点（限时 + 审计日志），用于 hotfix 等紧急场景 |
+| `/regress:characterize` | 特征测试生成：改无测试老代码前先钉住现状行为（golden master——探针真跑取真值，不猜输出） |
+| `/regress:evolve` | 遇到难点时查社区经验→适配项目→沉淀为知识库（不再重复查） |
+| `/regress:finish` | 收尾流水线：track 回写 → verify 全证据 → 代谢沉淀 → 状态推进 → 汇总报告（一条命令走完收尾，人只出现在决策点） |
+| `/regress:init` | 初始化项目的 .regress/ 数据目录（代码已在全局级安装，这里只建数据） |
+| `/regress:install` | 一键安装 regress-guard 到用户级（零配置，装一次全局生效） |
+| `/regress:learn` | 分析历史+检测框架规则，输出项目洞察，写入 AGENTS.md |
+| `/regress:plan` | 需求→解析→消歧→改动清单。AI 先补全上下文再动手（不问能推断的，只问关键分歧） |
+| `/regress:quick` | 快速模式——机器判据达标的小改动专用（合并 plan+track 一步到位） |
+| `/regress:resume` | 从 .regress/ 产物层单侧重建工作现场（断点续作，不依赖对话历史） |
+| `/regress:stats` | 健康报表：门禁拦截/债务/规律命中/僵尸清单/钩子活性一屏读（观测期仪表盘，只读不改） |
+| `/regress:trace` | 查看交付链：需求→会话→事件→提交 的可追溯视图（文本版 Inspector） |
+| `/regress:track` | 对比 git diff 发现 F3 并直接回写（AI 完成修改后自动执行，不需用户手动触发） |
+| `/regress:uninstall` | 卸载 regress-guard（清理用户级配置 + hook + skill + 命令） |
+| `/regress:update` | 检查并更新 regress-guard 到最新版（手动触发；SessionStart 也会自动检测） |
+| `/regress:verify` | 跑测试预览结果（提交时 hook 也会自动跑；测试失败 AI 应自行修复重跑） |
+
+<!-- generated: reference end -->