lingtai-dev-guide

Routes LingTai contributors to nested references for setup, debugging, security audits, and PR readiness.

692|60|Updated Mar 15, 2026
One-click install
npx skills add https://github.com/Lingtai-AI/lingtai --skill lingtai-dev-guide
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: lingtai-dev-guide
Source: https://github.com/Lingtai-AI/lingtai/tree/main/tui/internal/preset/skills/lingtai-dev-guide
Command: npx skills add https://github.com/Lingtai-AI/lingtai --skill lingtai-dev-guide

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) and scripts (resource) components.

What problem does it solve?

Contributing to the LingTai project spans a Go TUI, a Go/React portal, a Python kernel, and multiple MCP addon repos, and developers often do not know which procedure, build command, or diagnostic applies to their task. This Skill acts as a router that directs contributors to the correct nested reference before they touch code, preventing stale assumptions and unsafe side effects.

Core Features & Use Cases

  • Task routing: A routing table maps intents (set up a dev environment, diagnose a stuck network, audit secrets, prepare a PR) to the right nested reference under reference/<topic>/.
  • Non-negotiable rules: Enforces progressive disclosure, anatomy-travels-with-code, explicit human authorization gates for GitHub mutations, and PYTHONPATH hygiene for agent launches.
  • Nested references: Covers architecture, setup, contributing workflow, gotchas, debug-troubleshoot, security-audit, runtime self-check, PR review deliverables, skill stewardship, repo watch, and cache hit-rate measurement.
  • Use Case: A developer whose merged kernel fix is not live at runtime follows the router to reference/runtime-self-check/SKILL.md, probes the actual import path and git HEAD, refreshes, and verifies the fix with an in-situ probe.

Quick Start

Ask the agent to load the lingtai-dev-guide skill and route your current LingTai development task, such as setting up a dev environment or diagnosing a stuck agent network, to the correct nested reference.

Frequently Asked Questions about lingtai-dev-guide

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up a LingTai development environment?

Clone the lingtai Go monorepo and lingtai-kernel Python repo, build the TUI and portal with make build, symlink the binaries into /opt/homebrew/bin for dev mode, and install the kernel editable into the TUI runtime venv using uv pip install -e. The setup reference lists prerequisites including Go 1.21+, Node 18+, and Python 3.10+.

How do I verify which LingTai code is actually running after a refresh?

Probe the runtime venv Python (not PATH python) to print lingtai.__file__, the editable install status, and the git HEAD of the imported checkout. Then confirm the active binary symlink target and run an in-situ live probe, since imported source alone does not prove long-lived runtime objects were rebuilt.

Why does my merged kernel fix not take effect in the running agent?

The runtime venv may import an editable checkout that is behind origin/main, or a leftover PYTHONPATH may shadow the venv import. Fast-forward the imported checkout first, then refresh, and verify with a live probe because refresh re-inherits the parent environment verbatim.

Can the agent open or merge PRs on its own?

No. Opening, editing, or merging PRs, pushing commits, and filing or closing issues all require an explicit imperative human authorization for that specific side effect. Generating diffs, running read-only reviews, and writing local HTML explainers are always allowed.

How do I measure prompt-cache hit rate for LingTai agents?

Use the bundled read-only script scripts/cache_hit_rate.py against an agent workdir, project root, or single token_ledger.jsonl file. It computes sum(cached)/sum(input) over rolling windows (1h/5h/1d/3d) and avoids double-counting daemon ledger entries.

What are the limitations of this developer guide?

It is a router, not a complete manual: detailed procedures live in nested reference files that must be read before acting, and repository-local developer guidance takes precedence over this global guide. It is for contributors, not end users, who should use tutorial-guide instead.