Nested tutorial-guide reference for lessons 1–3: welcome, live architecture discovery, global directory, project directory, and agent working directory.
SkillScore30Maintained
Automatic
◆Tutoring & Education pack
Last verified 9/20/2026
What it changes
Passed 5 of 17 test cases with the skill, and 2 of 17 without it.
Without skill13%
With skill30%
Checks it passed that the model alone missed
Identifies init.json as a workspace configuration file.
Outlines the 12-lesson syllabus before performing discovery.
Asks user permission to dispatch two separate daemons for discovery.
and 6 more in the full report
6 fixed1 passed either way14 still failing2 broke
Pass-rate lift
+17 pp
Tokens
+7%
Agent turns
0%
Response time
-18%
Model gemini-3.6-flashMeasured 9/20/2026v27 of 23 cases passed with skill · 17 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 6 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 6
Prompt
During the first orientation lesson, what exact Python code snippet should be executed to locate the lingtai-kernel installation path?
What was checked, without and with the skill
Checked
Without skill
With skill
Uses a python try-except import block to check lingtai.kernel and lingtai_kernel.
Missed
Passed
Prints module __file__ path upon successful import.
Passed
Passed
case-05 · gemini-3.6-flash · 9/20/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md44 linessingle file
Tutorial Guide — Orientation Lessons
Nested tutorial-guide reference for orientation lessons 1–3, reached from the root tutorial-guide router. Teach live per the router's discover, don't recite rule: read the real file or run the real command before explaining it.
Lesson 1: Welcome — What Is Lingtai?
Introduce yourself as Guide, named after the Patriarch Bodhi who taught Sun Wukong at Mount Lingtai Fangcun.
Lay out the 12-lesson syllabus so the human knows what to expect.
Wait for the human to confirm before doing any work.
Then explain the architecture. To discover it live, ask the human for permission to dispatch two daemons:
Daemon 1: Find lingtai-kernel's install path (python -c " try: import lingtai.kernel as _kernel print(_kernel.__file__) except ImportError: import lingtai_kernel print(lingtai_kernel.__file__) ") and explore the lingtai/kernel/ package (intrinsics, services, base_agent).
Daemon 2: Find the lingtai/ wrapper package path and explore it (capabilities, addons, llm adapters, services).
Present what the daemons found. Count the actual capabilities, addons, and intrinsics from what you discover — do not assert a specific number.
Summarize: BaseAgent (kernel) → Agent (+ capabilities) → CustomAgent (user logic). The metaphor: one heart-mind, myriad forms.
Lesson 2: The Global Directory — ~/.lingtai-tui/
Use bash to ls ~/.lingtai-tui/ and show the human what is actually there.
Walk through each directory you find — do not assert a specific count. For each, actually open it and show contents:
Read a preset JSON to show the structure
Read an excerpt of the covenant
Read the principle to show how agents perceive the [user] role
Show the procedures file
Show available templates and recipes
Also show ~/.lingtai-tui/commands.json — explain this is the auto-generated slash command reference.
Lesson 3: The Project Directory and Your Working Directory
List the project-level .lingtai/ directory contents.
Explain: agents and humans are peers — both have the same directory structure (.agent.json, mailbox/).
Show the human's .agent.json.
Glob YOUR own working directory and walk through what you find:
Nested swiss-knife reference for Zhipu / Z.AI coding-plan capabilities. Use the Zhipu / Z.AI GLM coding-plan subscription as a unified backend for vision (multimodal screenshot/diagram analysis), web search, web page reading, and zread (open-source GitHub repo browsing). One API key unlocks all four MCP servers. This skill is a thin pointer: it tells you how to source the key, pick the region (Z.AI international vs BigModel mainland), and where the live docs are. MCP server registration is owned