refactor: Phase 1+2 — project-level output structure, 78→79 sync, gitignore fixes

.github/copilot-instructions.md

@@ -0,0 +1,31 @@
+# Copilot Instructions
+
+This is a Test-Agent project with a defined Skill → Agent → Utils execution framework.
+
+## When generating code or answering questions:
+
+1. **Read Skill docs first** — Check `skills/<task>.md` to understand the workflow before generating any test code or calling agents.
+2. **Follow the framework** — Do not skip the Skill workflow layer. Agent definitions are in `agents/`, utils in `utils/`.
+3. **Use existing utils** — The project has 79 utility modules in `utils/`. Prefer using them over generating new code.
+4. **Match the project structure** — Test outputs go to `workspace/`, config in `config/`, source code in `src/`.
+5. **Config first** — Check `.env.example` for required environment variables before generating code that depends on them.
+
+## Standalone CLI
+
+The project has a CLI at `runtime/cli/main.py` (entry point: `tagent`):
+```bash
+tagent run "path/to/prd.md"
+tagent catalog
+tagent doctor
+```
+
+## Key directories
+
+| What | Where |
+|------|-------|
+| Agent definitions | `agents/` |
+| Skill workflow docs | `skills/` |
+| Python utils | `utils/` |
+| Runtime (CLI + MCP) | `runtime/` |
+| Config templates | `config/` |
+| Test outputs | `workspace/` |

.github/workflows/ci.yml

@@ -293,7 +293,7 @@ jobs:
           echo "Agents=$AGENTS Skills=$SKILLS Utils=$UTILS"
           [ "$AGENTS" = "16" ] || { echo "❌ Agents count mismatch (expected 16, got $AGENTS)"; exit 1; }
           [ "$SKILLS" -eq "32" ] || { echo "❌ Skills count mismatch (expected 32, got $SKILLS)"; exit 1; }
-          [ "$UTILS" -eq "78" ] || { echo "❌ Utils count mismatch (expected 78, got $UTILS)"; exit 1; }
+          [ "$UTILS" -eq "79" ] || { echo "❌ Utils count mismatch (expected 79, got $UTILS)"; exit 1; }
           echo "✅ File counts correct"
 
       - name: 验证 .gitignore 排除源 MD

.gitignore

@@ -17,14 +17,57 @@ htmlcov/
 
 # workspace 测试产出（敏感 + 体积大）
 workspace/测试数据/
+workspace/automotive/
+workspace/测试用例/
+workspace/测试计划/
+workspace/需求分析/
 workspace/执行日志/
-workspace/测试报告/
+workspace/audit/
 workspace/_outputs/
 workspace/feedback/
 workspace/自动化脚本/
 
-# 但保留 workspace/执行日志/baselines/（性能基线需提交）
-!workspace/执行日志/baselines/
+# 测试报告 — 排除产物，保留 baselines/
+# 根级直接文件（临时报告、自检产物等）
+workspace/测试报告/*.md
+workspace/测试报告/*.docx
+workspace/测试报告/*.json
+# 旧式子目录（v1.42 前，无项目名层级）
+workspace/测试报告/allure-results/
+workspace/测试报告/allure-report/
+workspace/测试报告/jmeter-results/
+workspace/测试报告/jmeter-report/
+workspace/测试报告/decisions/
+workspace/测试报告/evidence/
+workspace/测试报告/reports/
+workspace/测试报告/security/
+workspace/测试报告/history/
+workspace/测试报告/quarantine/
+workspace/测试报告/soak/
+workspace/测试报告/ai-*/
+workspace/测试报告/web-vitals/
+workspace/测试报告/silent-failures/
+workspace/测试报告/audit/
+workspace/测试报告/screenshots/
+# 按 run_id 隔离的产出目录（时间戳格式: YYYYMMDD_HHMMSS_xxxxxx）
+workspace/测试报告/*/2*/
+# 项目级产出子目录（{项目名}/ 下）
+workspace/测试报告/*/screenshots/
+workspace/测试报告/*/allure-results/
+workspace/测试报告/*/allure-report/
+workspace/测试报告/*/jmeter-results/
+workspace/测试报告/*/jmeter-report/
+workspace/测试报告/*/decisions/
+workspace/测试报告/*/evidence/
+workspace/测试报告/*/reports/
+workspace/测试报告/*/security/
+workspace/测试报告/*/history/
+workspace/测试报告/*/quarantine/
+workspace/测试报告/*/soak/
+workspace/测试报告/*/ai-*/
+workspace/测试报告/*/web-vitals/
+workspace/测试报告/*/silent-failures/
+workspace/测试报告/*/audit/
 
 # ===== 敏感配置 =====
 .env
@@ -41,6 +84,8 @@ workspace/自动化脚本/
 Thumbs.db
 
 # ===== Claude Code =====
+.claude/agents/
+.claude/skills/
 .claude/cache/
 .claude/sessions/
 

.pre-commit-config.yaml

@@ -67,7 +67,7 @@ repos:
       # 文件统计校验（防误删 agent/skill/utils）
       - id: file-count-check
         name: Agents/Skills/Utils 数量校验
-        entry: bash -c 'A=$(ls agents/[0-9]*.md 2>/dev/null | wc -l); S=$(ls skills/*.md 2>/dev/null | grep -v README | wc -l); U=$(find utils -name "*.py" ! -name "__init__.py" 2>/dev/null | wc -l); [ "$A" = "16" ] || { echo "❌ Agents count mismatch (expected 16, got $A)"; exit 1; }; [ "$S" -eq "32" ] || { echo "❌ Skills count mismatch (expected 32, got $S)"; exit 1; }; [ "$U" -eq "78" ] || { echo "❌ Utils count mismatch (expected 78, got $U)"; exit 1; }; echo "✅ File counts correct"'
+        entry: bash -c 'A=$(ls agents/[0-9]*.md 2>/dev/null | wc -l); S=$(ls skills/*.md 2>/dev/null | grep -v README | wc -l); U=$(find utils -name "*.py" ! -name "__init__.py" 2>/dev/null | wc -l); [ "$A" = "16" ] || { echo "❌ Agents count mismatch (expected 16, got $A)"; exit 1; }; [ "$S" -eq "32" ] || { echo "❌ Skills count mismatch (expected 32, got $S)"; exit 1; }; [ "$U" -eq "79" ] || { echo "❌ Utils count mismatch (expected 79, got $U)"; exit 1; }; echo "✅ File counts correct"'
         language: system
         pass_filenames: false
         always_run: true

AGENTS.md

@@ -0,0 +1,78 @@
+# AGENTS.md
+
+## AI AGENT — STOP. READ THIS FIRST.
+
+**This is a Test-Agent project. It has a defined execution framework. You MUST follow it.**
+
+- Do NOT call or spawn sub-agents without reading the relevant Skill doc first.
+- Do NOT skip the Skill workflow layer.
+- Do NOT execute test workflows directly — read the Skill doc and follow its agent call sequence.
+
+**Correct flow:**
+1. Read `skills/<task>.md` to understand the workflow
+2. Follow the Skill doc's step-by-step agent call sequence
+3. Use Agent tools ONLY when and how the Skill doc says
+
+**If no Skill doc matches the task:** fall back to `skills/test-coordinator.md`.
+
+---
+
+## Two Ways to Use Test-Agent
+
+### Way 1: Standalone CLI (no AI required)
+
+```bash
+cd project-dir
+pip install -e runtime/          # first time only
+tagent run "path/to/prd.md"      # router + orchestrator end-to-end
+tagent catalog                   # list 16 experts + 32 skills
+tagent doctor                    # health check
+```
+
+### Way 2: AI Agent Collaboration
+
+```bash
+cd project-dir
+# Read skills/smoke-test.md first, then follow workflow
+# Call sub-agents per Skill doc sequence (requirements-analyst → testcase-designer → ...)
+# Outputs in workspace/
+```
+
+---
+
+## Architecture
+
+```
+Standalone CLI (tagent)              AI Agent Mode
+        │                                      │
+        ▼                                      ▼
+   runtime/router                    Skill docs (skills/*.md)
+        │                                      │
+        ▼                                      ▼
+   runtime/orchestrator              Agent defs (agents/*.md)
+        │                                      │
+        ▼                                      ▼
+   utils/*.py (79 modules)           utils/*.py (79 modules)
+```
+
+Both paths converge at the utils execution layer.
+
+## Directory Map
+
+| What | Where |
+|------|-------|
+| Agent definitions | `agents/` (16 agents) |
+| Skill workflow docs | `skills/` (35 skills) |
+| Python utils | `utils/` (79 modules) |
+| Runtime (CLI + orchestrator + MCP) | `runtime/` |
+| Config templates | `config/` (incl. `.env.example`) |
+| Test outputs | `workspace/` |
+| CI pipelines | `ci/` |
+| Docs | `docs/` |
+
+## Design Notes
+
+- Slash commands like `/smoke-test` are Skill workflow documents, not registered commands.
+- Agent definitions contain Python import hints as prompts for AI agents — not actual executable code.
+- Pentest workflows require `tagent.yml` with `pentest.authorized: true`.
+- MCP server (`python -m runtime.mcp.test_orchestrator.server`) provides catalog/plan/run/status/report tools.

CHANGELOG.md

@@ -320,8 +320,8 @@ _后续累积变更入此节;切版本时移到下方版本节。_
   - `workspace/需求分析/requirements_summary.json`(需求摘要 + P0/P1 + 风险区)
   - `workspace/测试计划/automation_scripts_plan.json`(脚本规划 + fixture 复用)
   - `workspace/测试计划/execution_plan.json`(4 阶段 + 失败 4 分类 + Flaky 规则)
-  - `workspace/测试报告/bug_drafts.json`(BugTracker-ready Bug 草案 + severity 1-4)
-  - `workspace/测试报告/decisions/final_verdict_*.json`(test-lead go/no-go 决策,标 `requires_human_signoff: true`)
+  - `workspace/测试报告/{项目名}/bug_drafts.json`(BugTracker-ready Bug 草案 + severity 1-4)
+  - `workspace/测试报告/{项目名}/decisions/final_verdict_*.json`(test-lead go/no-go 决策,标 `requires_human_signoff: true`)
 - **录制脚本**(`scripts/`):
   - `_demo-commands.sh`:实际 demo 命令序列(被 record-demo-* 调)
   - `record-demo-asciinema.sh`:`asciinema rec` 自动录,产 .cast 可上传 asciinema.org 或转 GIF/SVG
@@ -382,7 +382,7 @@ _后续累积变更入此节;切版本时移到下方版本节。_
   - `00-项目导航.md` · `agents/{01,07,08,09}.md` · `agents/README.md` · `skills/{README,test-coordinator,zentao-bug-submission}.md` · `config/mcp-server-impl.md` · `utils/{README.md,api_retry_util.py}` · `ci/{INDEX,CICD集成说明}.md` · `docs/getting-started/{交付物清单,使用手册,配置清单}.md` · `examples/web-demo/README.md` · `CONTRIBUTING.md` · `FULL_GUIDE.md`
 - **adapter 修 V1.10 n7 bug**:`runtime/orchestrator/adapters/experts.py` 加 `SCRIPT_DEFAULT_ARGS` + `_ensure_fixture()` 通用机制
   - 现 `tagent selftest --e2e --strict` **100% PASS 8/8**(原 88% 7/8)
-  - generate_report.py 默认注入 `--data=workspace/测试报告/_selftest_summary.json`,fixture 自动生成
+  - generate_report.py 默认注入 `--data=workspace/测试报告/{项目名}/_selftest_summary.json`,fixture 自动生成
 - **主宪章扩**:
   - §36 多端通知 canon(扩 §6,6 渠道权威清单 + env 字段 + 业务语言铁律)
   - §37 BugTracker canon(扩 §12,6 adapter 权威清单 + measurement env + 措辞规范)
@@ -542,7 +542,7 @@ _后续累积变更入此节;切版本时移到下方版本节。_
   - `runtime/mcp/defect_tracker/`:工单桥 5 工具(create/get/update/query_bugs/list_trackers),默认 zentao + 预留扩展位(主宪章 §12 契约)
   - `runtime/mcp/knowledge_base/`:pgvector 向量检索 4 工具(embed/index_case/index_defect/search_similar),LiteLLM embedding + stub 兜底
   - `runtime/mcp/compliance_checker/`:行业合规规则 3 工具(list_profiles/get_profile/check_compliance);10 框架 profile 起步空载(SOC2/PCI-DSS/HIPAA/IEC 62304/IEC 61508/ISO 26262/DO-178C/GDPR/PIPL/CCPA)
-  - 共享基类 `runtime/mcp/base.py`:make_server / run_stdio / @tool_decision_logged(决策落 `workspace/测试报告/decisions/` 符合主宪章 §18-12)
+  - 共享基类 `runtime/mcp/base.py`:make_server / run_stdio / @tool_decision_logged(决策落 `workspace/测试报告/{项目名}/decisions/` 符合主宪章 §18-12)
 - **行业合规规则插槽** `profiles/compliance/`:10 框架空载示例 YAML,真规则由领域专家+test-lead 双签签字后入库
 - **飞轮回灌路由**(M2-9):`runtime/router/retrieval.py` 历史相似用例 → LLM prompt few-shot;router 透明集成,无 KB 时降级
 - **真模型路由测试套件**(M2-7):`runtime/tests/test_router_real.py` 20 样本(4 类型 × 5)真模型测试;门槛单模型 ≥85%、双模型投票 ≥95%;无 API key 自动 skip;失败自动落 decisions/ 含 seed+模型版本+输入快照(主宪章 §21 横切准则)