القائمة
Use when testing NanoHermes AI Agent system — provides progressive-disclosure testing methodology with 131 test cases across 15 core modules, advanced scenarios, and report templates.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
| name | nanohemes-testing-guide |
| description | Use when testing NanoHermes AI Agent system — provides progressive-disclosure testing methodology with 131 test cases across 15 core modules, advanced scenarios, and report templates. |
| version | 1.0.0 |
| author | Hermes Agent |
| license | MIT |
| metadata | {"hermes":{"tags":["testing","nanohemes","qa","e2e","progressive-disclosure"],"related_skills":["writing-plans","requesting-code-review"]}} |
Complete end-to-end testing methodology for the NanoHermes AI Agent system, covering 15 core modules with 131 test cases, 36 advanced scenarios, and reusable report templates. Designed for progressive disclosure — load only the detail you need.
Target project: /mnt/d/code/NanoHermes (Python self-evolving AI Agent, ~960 tests, 15 modules)
Don't use for: Unit-level pytest runs (use python -m pytest tests/ directly), or testing unrelated Python projects.
The testing process follows 10 sequential steps, each validating a different system layer:
| Step | Focus | Key Commands | Expected |
|---|---|---|---|
| 1 | 启动验证 | python -m src.main | 配置加载、工具注册、记忆注入 |
| 2 | 基础对话 | 发送"你好" | 响应 <10 秒，流式输出正常 |
| 3 | 工具系统 | 测试 terminal/read_file/write_file/patch/search_files/execute_code/clarify | 12 个工具全部可用 |
| 4 | 会话存储 | 检查 ~/.nanohermes/sessions/ | JSONL 文件 + SQLite 元数据 |
| 5 | 记忆系统 | 检查 ~/.nanohermes/memory/ | MEMORY.md/USER.md 正确注入 |
| 6 | Provider 运行 | 观察 API 调用 | DashScope qwen3.6-plus 正常响应 |
| 7 | 对话循环 | 上下文保持、错误恢复、多工具调用 | 上下文理解准确，错误不崩溃 |
| 8 | 配置系统 | 验证优先级链：显式 > JSON > .env > 默认 | 配置正确加载 |
| 9 | TUI 界面 | 界面渲染、命令处理、键盘交互 | 流式输出、Tab 补全正常 |
| 10 | 边界情况 | 空输入、超长输入、特殊字符 | 优雅处理，无崩溃 |
Environment setup (run before testing):
cd /mnt/d/code/NanoHermes
eval "$($HOME/miniconda3/bin/conda shell.bash hook)" && conda activate py312
# Verify: .env has DASHSCOPE_API_KEY, DASHSCOPE_BASE_URL, MODEL_NAME
This skill uses progressive disclosure to avoid overwhelming context. Start with this SKILL.md, then load specific reference files as needed:
skill_view(name='nanohemes-testing-guide', file_path='references/test-cases.md')
→ 131 test cases across 15 modules with checkable status boxes
skill_view(name='nanohemes-testing-guide', file_path='references/advanced-scenarios.md')
→ 36 advanced scenarios: tool combinations, session management, memory conflicts, error injection, performance benchmarks, security testing
skill_view(name='nanohemes-testing-guide', file_path='references/troubleshooting.md')
→ Known issues, failure patterns, debug procedures, performance metrics
skill_view(name='nanohemes-testing-guide', file_path='references/pty-test-results.md')
→ Real PTY test execution results: 23/23 pass, performance metrics, SQLite schema verified
skill_view(name='nanohemes-testing-guide', file_path='references/test-artifacts-convention.md')
→ testing-artifacts/ directory rules: never overwrite reports, auto-rename on collision, script reuse first
skill_view(name='nanohemes-testing-guide', file_path='references/report-template.md')
→ Reusable markdown test report template. Must save to testing-artifacts/reports/report-YYYY-MM-DD-HHMM.md — never overwrite existing reports; auto-append sequence number on collision.
skill_view(name='nanohemes-testing-guide', file_path='references/sqlite-counter-fix.md')
→ SQLite counter debugging guide: how to diagnose and fix message_count/tool_call_count always being 0.
| Module | Test Cases | Status |
|---|---|---|
| session-storage | 7 | ⬜ |
| tool-runtime | 12 | ⬜ |
| memory-system | 5 | ⬜ |
| provider-runtime | 6 | ⬜ |
| unified-config-system | 7 | ⬜ |
| system-prompt-assembly | 8 | ⬜ |
| tool-search | 6 | ⬜ |
| conversation-loop | 10 | ⬜ |
| multi-agent-delegation | 5 | ⬜ |
| skill-system | 6 | ⬜ |
| context-compression | 5 | ⬜ |
| insights-metrics | 4 | ⬜ |
| add-mcp-server-support | 6 | ⬜ |
| cli (TUI) | 8 | ⬜ |
| auxiliary | 4 | ⬜ |
| Advanced scenarios | 36 | ⬜ |
| TOTAL | 167 | ⬜ |
| Resource | Path |
|---|---|
| Project root | /mnt/d/code/NanoHermes |
| Session storage | ~/.nanohermes/sessions/ (JSONL) |
| Session metadata | ~/.nanohermes/sessions.db (SQLite) |
| Memory files | ~/.nanohermes/memory/MEMORY.md, USER.md |
| MCP config | ~/.nanohermes/mcp_servers.json |
| Environment | .env (DASHSCOPE_API_KEY, DASHSCOPE_BASE_URL, MODEL_NAME) |
| Test outputs | testing-artifacts/{reports,scripts,logs} (all test artifacts go here) |
| Test guide doc | TESTING_GUIDE.md (project root, full 131-case guide) |
# Run all tests
python -m pytest tests/ -v
# Run module tests
python -m pytest tests/tools/ -v
python -m pytest tests/provider/ -v
# E2E test (interactive)
python -m src.main # normal mode
python -m src.main --debug # debug mode (full JSON request/response)
python -m src.main --resume # resume last session
# Clean test data
rm -rf ~/.nanohermes/sessions/*
rm -f ~/.nanohermes/sessions.db
rm -rf ~/.nanohermes/memory/*
# Verify storage
ls -la ~/.nanohermes/sessions/
sqlite3 ~/.nanohermes/sessions.db "SELECT COUNT(*) FROM sessions;"
cat ~/.nanohermes/memory/MEMORY.md
memory tool./tools command. Impact: low (efficiency). Fix: narrow intent in prompt.--debug mode to slow down.~/.nanohermes/sessions/ before each test run.testing-artifacts/ (created by mkdir -p testing-artifacts/{reports,scripts,logs}). Never leave test output in project root. See references/test-artifacts-convention.md for conflict resolution rules.src/conversation/loop.py constructor. TUI (src/cli/tui.py) creates ConversationLoop without passing max_iterations, so it always uses 90. No config-level override exists. If you need more iterations for complex tool-chain tests, either patch the TUI to pass it, or modify the default in loop.py.process(action='poll') to get output, allow 3-15s sleep between submits for LLM response.message_count / tool_call_count in sessions table were never incremented. Fixed: session_db.py:insert_message auto-calls increment_message_count for user/assistant/system roles; event_handler.py:_on_tool_start calls increment_tool_call_count. If you find counters at 0 again, check these two call sites.__init__ accepts session_db, jsonl_store, memory_manager, skill_manager keyword args using _UNSET sentinel pattern. _UNSET = auto-initialize, None = disabled, instance = use injected. Tests that mock storage must pass these as kwargs: TUIApp(session_db=my_db, jsonl_store=my_store).tests/tools/test_integration.py imports src.tools.toolsets which doesn't exist. Skip with --ignore=tests/tools/test_integration.py until the module is created or the test is updated.testing-artifacts/logs/fix-log.md.After running tests:
references/test-cases.md for checklist)references/advanced-scenarios.md)references/report-template.md)Integration testing for AI Agent CLI systems — PTY-driven conversation simulation, tool chain verification, OpenSpec-driven test design, and real-API validation.
Extract web article content from WeChat MP, blogs, and other platforms, converting text, images, formatting, and tables to clean Markdown.
Configure, extend, or contribute to Hermes Agent.
Generate images, video, and audio with ComfyUI — install, launch, manage nodes/models, run workflows with parameter injection. Uses the official comfy-cli for lifecycle and direct REST/WebSocket API for execution.
Decomposition playbook + anti-temptation rules for an orchestrator profile routing work through Kanban. The "don't do the work yourself" rule and the basic lifecycle are auto-injected into every kanban worker's system prompt; this skill is the deeper playbook when you're specifically playing the orchestrator role.
Pitfalls, examples, and edge cases for Hermes Kanban workers. The lifecycle itself is auto-injected into every worker's system prompt as KANBAN_GUIDANCE (from agent/prompt_builder.py); this skill is what you load when you want deeper detail on specific scenarios.