lingtai-skill
May 30, 2026 · View on GitHub
The canonical skill describing how to interact with a LingTai agent network through the shared human mailbox.
This is a single Markdown file at skills/lingtai/SKILL.md documenting the filesystem mailbox protocol — reading and sending mail, agent discovery, liveness checks, lifecycle management, and signal files. The protocol is pure filesystem; no SDK, no API, no dependencies.
Who this is for
Coding agents that can read and write files. The skill is consumed by:
- Lingtai-AI/claude-code-plugin — Claude Code plugin (auto-loads via
claude plugin add) - Lingtai-AI/codex-plugin — OpenAI Codex CLI plugin
- Any other coding agent with file access — clone and copy
skills/lingtai/SKILL.mdinto your tool's skill directory
Manual install
For agents without a dedicated wrapper plugin:
git clone https://github.com/Lingtai-AI/lingtai-skill.git
cp -r lingtai-skill/skills/lingtai <your-tool's-skill-dir>/
For example:
- OpenCode:
~/.config/opencode/skills/lingtai/ - A tool that reads
AGENTS.md:cat lingtai-skill/skills/lingtai/SKILL.md >> AGENTS.md
Why this lives in its own repo
The skill describes a protocol, not a tool. Both the Claude Code plugin and the Codex CLI plugin consume the same SKILL.md verbatim — keeping it here means there is exactly one source of truth, and the host-specific plugin repos own only their host-specific glue (hooks, manifests, install scripts).
Requirements
- A running LingTai project (
.lingtai/directory with agents) - File read/write access
- Python 3 (for UUID generation and liveness checks)
License
Apache-2.0