Create Agent QC v0.1.0 standard

Author: wutongci

.github/workflows/pages.yml

@@ -0,0 +1,55 @@
+name: Deploy documentation to GitHub Pages
+
+on:
+  push:
+    branches: [main]
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  pages: write
+  id-token: write
+
+concurrency:
+  group: pages
+  cancel-in-progress: false
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup Node
+        uses: actions/setup-node@v4
+        with:
+          node-version: 22
+          cache: npm
+
+      - name: Install dependencies
+        run: npm ci
+
+      - name: Validate examples
+        run: npm run check:schemas
+
+      - name: Build site
+        run: npm run build
+        env:
+          VITEPRESS_BASE: "/${{ github.event.repository.name }}/"
+
+      - name: Upload artifact
+        uses: actions/upload-pages-artifact@v3
+        with:
+          path: docs/.vitepress/dist
+
+  deploy:
+    environment:
+      name: github-pages
+      url: ${{ steps.deployment.outputs.page_url }}
+    runs-on: ubuntu-latest
+    needs: build
+    steps:
+      - name: Deploy to GitHub Pages
+        id: deployment
+        uses: actions/deploy-pages@v4

.gitignore

@@ -0,0 +1,5 @@
+node_modules/
+docs/.vitepress/dist/
+.DS_Store
+*.log
+!package-lock.json

CHANGELOG.md

@@ -0,0 +1,11 @@
+# Changelog
+
+## v0.1.0 - 2026-05-10
+
+- Initial Lime-focused Agent QC draft standard.
+- Added QC object model for plans, cases, gates, runs, verdicts, evidence, and reports.
+- Added Lime gate matrix for local checks, contract tests, GUI smoke, Playwright, qcloop, and review.
+- Added qcloop integration rules and public JSON examples.
+- Added public schemas for `qc_plan`, `qc_case`, and `qc_verdict`.
+- Added bilingual documentation and LLM entrypoints.
+- Added GitHub Pages deployment workflow for the documentation site.

README.md

@@ -0,0 +1,77 @@
+# Agent QC
+
+Agent QC is a Lime-focused draft standard for evidence-driven testing plans, quality gates, qcloop batches, verdicts, and review reports.
+
+Lime is a GUI desktop product. For Lime, "tests passed" is not enough unless the result proves the relevant product path: local checks, command contracts, bridge readiness, GUI smoke, Playwright interaction, qcloop batch results, and human or LLM judge evidence all need a shared language.
+
+Agent QC defines that language.
+
+## Core boundary
+
+| System | Owns | Agent QC owns |
+| --- | --- | --- |
+| Lime source repo | Product code, scripts, runtime commands, GUI, docs, and mocks. | Which Lime gates are required for a change type. |
+| qcloop | Batch execution, worker/verifier/repair loop, SQLite state, attempts, and qc rounds. | How tests are planned, itemized, judged, and reported. |
+| Agent Evidence | Durable traces, provenance, review, replay, and export facts. | Evidence references required by QC verdicts. |
+| Agent UI | How testing tasks and evidence are projected to users. | Behavior-level acceptance scenarios and honest pass/fail semantics. |
+| Human reviewers | Final product judgment when needed. | How manual review decisions become explicit verdict evidence. |
+
+## What v0.1.0 defines
+
+- `qc_plan`: a Lime test plan with change type, risk, required gates, cases, and evidence policy.
+- `qc_case`: one testable item that can map to a qcloop item.
+- `qc_gate`: a required validation boundary such as `verify:local`, `test:contracts`, or `verify:gui-smoke`.
+- `qc_run`: one execution record with command, executor, environment, and output references.
+- `qc_verdict`: pass/fail judgment backed by evidence refs, not agent prose.
+- `qc_evidence`: references to command logs, GUI smoke output, Playwright traces, qcloop attempts, screenshots, reports, or review notes.
+- `qc_report`: final aggregate result with remaining risk and next-action guidance.
+
+## Lime testing stance
+
+Agent QC uses evidence-driven gates:
+
+1. Pick the Lime change type first.
+2. Select the minimum required gates from the Lime gate matrix.
+3. Write behavior-level `qc_case` items.
+4. Execute small deterministic gates directly and batch repetitive cases through qcloop.
+5. Let an independent verifier judge evidence using strict JSON.
+6. Report pass/fail/exhausted with evidence refs and remaining risk.
+
+A qcloop batch finishing is not automatically a product pass. A Lime change is deliverable only when the required gates for its risk class have evidence-backed verdicts.
+
+## Documentation
+
+- [Specification](docs/en/specification.md)
+- [Lime gate matrix](docs/en/authoring/lime-gate-matrix.md)
+- [qcloop integration](docs/en/authoring/qcloop-integration.md)
+- [Evidence-driven verdicts](docs/en/authoring/evidence-driven-verdicts.md)
+- [Acceptance scenarios](docs/en/authoring/acceptance-scenarios.md)
+- [中文规范](docs/zh/specification.md)
+
+## LLM entrypoints
+
+- [`llms.txt`](llms.txt): concise navigation index for AI clients.
+- [`llms-full.txt`](llms-full.txt): concatenated current English documentation for model context.
+- [`llm.txt`](llm.txt) and [`llm-full.txt`](llm-full.txt): compatibility aliases.
+
+## Local development
+
+```bash
+npm install
+npm run dev
+```
+
+## Build and validation
+
+```bash
+npm run check:schemas
+npm run build
+```
+
+The static site is generated at `docs/.vitepress/dist`.
+
+## Related local sources
+
+- Lime engineering rules: `/Users/coso/Documents/dev/ai/aiclientproxy/lime/docs/aiprompts/quality-workflow.md`
+- Lime GUI E2E rules: `/Users/coso/Documents/dev/ai/aiclientproxy/lime/docs/aiprompts/playwright-e2e.md`
+- qcloop reference: `/Users/coso/Documents/dev/ai/limecloud/qcloop/llms-full.txt`

RELEASE_NOTES.md

@@ -0,0 +1,38 @@
+# Release Notes
+
+## v0.1.0
+
+Agent QC v0.1.0 is the first Lime-focused draft standard for evidence-driven testing plans, quality gates, qcloop batches, verdicts, and release-ready reports.
+
+### Highlights
+
+- Defines the core Agent QC objects: `qc_plan`, `qc_case`, `qc_gate`, `qc_run`, `qc_verdict`, `qc_evidence`, and `qc_report`.
+- Establishes Lime's evidence-driven testing stance: a pass verdict requires inspectable evidence, not agent self-reporting.
+- Adds the Lime gate matrix for `verify:local`, `test:contracts`, `verify:gui-smoke`, Playwright product flows, qcloop batches, and manual or LLM judge review.
+- Documents qcloop integration, including JSON item shape, worker prompt requirements, verifier prompt requirements, and status mapping for `success`, `failed`, and `exhausted`.
+- Adds bilingual English and Simplified Chinese documentation with VitePress navigation, version pages, and LLM-friendly entrypoints.
+- Adds public JSON schemas and examples for QC plans, QC cases, QC verdicts, and qcloop job requests.
+- Adds GitHub Pages deployment through GitHub Actions.
+
+### Public docs
+
+After the Pages workflow completes, the documentation site is expected at:
+
+- https://limecloud.github.io/agentqc/
+- https://limecloud.github.io/agentqc/llms.txt
+- https://limecloud.github.io/agentqc/llms-full.txt
+
+### Validation
+
+Local validation performed before release:
+
+- `npm run check:schemas`
+- `npm run build`
+- `npm audit --omit=dev`
+- `npm audit --audit-level=high`
+
+`npm audit --audit-level=high` passes with no high or critical vulnerabilities. npm still reports a dev-only moderate advisory in the VitePress/Vite/esbuild chain with no current fix available.
+
+### Scope
+
+This release is intentionally scoped to Lime. It does not claim to be a generic industry testing standard yet. qcloop remains the execution loop; Agent QC defines the testing standard, evidence requirements, and reporting contract.

docs/.vitepress/config.ts

@@ -0,0 +1,133 @@
+import { defineConfig } from 'vitepress'
+
+const base = process.env.VITEPRESS_BASE || '/'
+const versions = ['0.1.0']
+
+const enVersionItems = versions.flatMap((version) => [
+  { text: `v${version} overview`, link: `/en/versions/v${version}/overview` },
+  { text: `v${version} specification`, link: `/en/versions/v${version}/specification` },
+  { text: `v${version} changelog`, link: `/en/versions/v${version}/changelog` }
+])
+
+const zhVersionItems = versions.flatMap((version) => [
+  { text: `v${version} 概览`, link: `/zh/versions/v${version}/overview` },
+  { text: `v${version} 规范`, link: `/zh/versions/v${version}/specification` },
+  { text: `v${version} 变更记录`, link: `/zh/versions/v${version}/changelog` }
+])
+
+const enNav = [
+  { text: 'Guide', link: '/en/what-is-agent-qc' },
+  { text: 'Specification', link: '/en/specification' },
+  { text: 'Examples', link: '/en/examples/lime-gui-change' },
+  { text: 'Ecosystem', link: '/en/reference/agent-ecosystem' },
+  { text: 'Version', items: [{ text: 'latest', link: '/en/specification' }, ...enVersionItems] }
+]
+
+const zhNav = [
+  { text: '指南', link: '/zh/what-is-agent-qc' },
+  { text: '规范', link: '/zh/specification' },
+  { text: '示例', link: '/zh/examples/lime-gui-change' },
+  { text: '生态', link: '/zh/reference/agent-ecosystem' },
+  { text: '版本', items: [{ text: 'latest', link: '/zh/specification' }, ...zhVersionItems] }
+]
+
+const enSidebar = [
+  {
+    text: 'Start here',
+    items: [
+      { text: 'Overview', link: '/en/' },
+      { text: 'What is Agent QC?', link: '/en/what-is-agent-qc' },
+      { text: 'Specification', link: '/en/specification' }
+    ]
+  },
+  {
+    text: 'For test authors',
+    items: [
+      { text: 'Quickstart', link: '/en/authoring/quickstart' },
+      { text: 'Lime gate matrix', link: '/en/authoring/lime-gate-matrix' },
+      { text: 'qcloop integration', link: '/en/authoring/qcloop-integration' },
+      { text: 'Evidence-driven verdicts', link: '/en/authoring/evidence-driven-verdicts' },
+      { text: 'Acceptance scenarios', link: '/en/authoring/acceptance-scenarios' }
+    ]
+  },
+  {
+    text: 'Reference',
+    items: [
+      { text: 'Glossary', link: '/en/reference/glossary' },
+      { text: 'Lime quality source index', link: '/en/reference/lime-quality-source-index' },
+      { text: 'Agent standards ecosystem', link: '/en/reference/agent-ecosystem' }
+    ]
+  },
+  {
+    text: 'Examples',
+    items: [
+      { text: 'Lime GUI change', link: '/en/examples/lime-gui-change' },
+      { text: 'Lime command bridge change', link: '/en/examples/lime-command-bridge-change' },
+      { text: 'qcloop batch', link: '/en/examples/qcloop-batch' }
+    ]
+  },
+  { text: 'Versions', items: enVersionItems }
+]
+
+const zhSidebar = [
+  {
+    text: '开始',
+    items: [
+      { text: '概览', link: '/zh/' },
+      { text: '什么是 Agent QC', link: '/zh/what-is-agent-qc' },
+      { text: '规范', link: '/zh/specification' }
+    ]
+  },
+  {
+    text: '测试作者',
+    items: [
+      { text: '快速开始', link: '/zh/authoring/quickstart' },
+      { text: 'Lime 门禁矩阵', link: '/zh/authoring/lime-gate-matrix' },
+      { text: 'qcloop 集成', link: '/zh/authoring/qcloop-integration' },
+      { text: '证据驱动判定', link: '/zh/authoring/evidence-driven-verdicts' },
+      { text: '验收场景', link: '/zh/authoring/acceptance-scenarios' }
+    ]
+  },
+  {
+    text: '参考',
+    items: [
+      { text: '术语表', link: '/zh/reference/glossary' },
+      { text: 'Lime 质量事实源索引', link: '/zh/reference/lime-quality-source-index' },
+      { text: 'Agent 标准生态', link: '/zh/reference/agent-ecosystem' }
+    ]
+  },
+  {
+    text: '示例',
+    items: [
+      { text: 'Lime GUI 改动', link: '/zh/examples/lime-gui-change' },
+      { text: 'Lime 命令桥接改动', link: '/zh/examples/lime-command-bridge-change' },
+      { text: 'qcloop 批次', link: '/zh/examples/qcloop-batch' }
+    ]
+  },
+  { text: '版本', items: zhVersionItems }
+]
+
+export default defineConfig({
+  base,
+  title: 'Agent QC',
+  description: 'A Lime-focused standard for evidence-driven agent quality control.',
+  cleanUrls: true,
+  lastUpdated: true,
+  themeConfig: {
+    logo: '/logo.svg',
+    nav: [
+      { text: 'English', items: enNav },
+      { text: '中文', items: zhNav }
+    ],
+    sidebar: {
+      '/en/': enSidebar,
+      '/zh/': zhSidebar
+    },
+    search: { provider: 'local' },
+    footer: {
+      message: 'Draft Lime-focused standard for evidence-driven agent quality control.',
+      copyright: 'Copyright © 2026'
+    }
+  },
+  markdown: { lineNumbers: true }
+})

docs/en/authoring/acceptance-scenarios.md

@@ -0,0 +1,61 @@
+---
+title: Acceptance scenarios
+description: Behavior-level validation scenarios for Lime Agent QC.
+---
+
+# Acceptance scenarios
+
+Agent QC scenarios validate behavior and evidence, not file presence.
+
+## 1. Ordinary frontend change
+
+1. Classify change as `frontend`.
+2. Run `npm run verify:local`.
+3. Record command output evidence.
+4. Report pass only if the command passes and no higher-risk boundary was touched.
+
+Passing condition: local validation evidence supports the change and no GUI or bridge gate is required.
+
+## 2. User-visible UI change
+
+1. Classify change as `user-visible-ui`.
+2. Run `npm run verify:local`.
+3. Add or run a stable UI assertion when an existing test surface exists.
+4. Record evidence for user-visible expected state.
+
+Passing condition: UI behavior is asserted, not just typechecked.
+
+## 3. GUI shell or Workspace change
+
+1. Classify change as `gui-shell-workspace`.
+2. Run `npm run verify:local`.
+3. Run `npm run verify:gui-smoke`.
+4. Record DevBridge and workspace readiness evidence.
+
+Passing condition: GUI main path is proven usable.
+
+## 4. Tauri command / Bridge / mock change
+
+1. Classify change as `tauri-command-bridge-mock`.
+2. Check command surfaces as a group.
+3. Run `npm run test:contracts` and `npm run verify:local`.
+4. Record failing command ids when contracts fail.
+
+Passing condition: frontend calls, Rust registration, catalog, and mocks stay synchronized.
+
+## 5. qcloop batch regression
+
+1. Convert each behavior into one independent `qc_case`.
+2. Create a qcloop job with strict verifier JSON.
+3. Start the job and poll until no item is pending or running.
+4. Summarize `success`, `failed`, and `exhausted` separately.
+
+Passing condition: every required item has verifier-backed evidence.
+
+## 6. Manual or LLM judge review
+
+1. Define a rubric before review.
+2. Provide the reviewer with outputs and evidence refs.
+3. Record verdict, reviewer identity or judge configuration, and remaining risk.
+
+Passing condition: review decision is explicit and traceable.