zhiping0913

Electrodynamics reasoning and knowledge graph — extends landau-graph with boundary-value methods, waveguides, antennas, scattering, crystal optics, coherence theory, and nonlinear/strong-field electrodynamics from 6 standard textbooks.

Optics engineering skill — extends landau-graph, electrodynamics, and plasma with Fourier optics, laser engineering, ultrashort pulse generation/measurement, acousto/electro-optics, stimulated scattering, and wave propagation in real media from 11 specialized textbooks.

Plasma physics reasoning and knowledge graph — extends landau-graph and electrodynamics with magnetized plasma waves, MHD stability, transport, laser-plasma interaction, and surface plasmonics from 10 standard textbooks.

LandauGraph — reasoning-constrained skill graph distilled from Landau-Lifshitz 10-volume Course of Theoretical Physics. 97 nodes, 196 edges. 4 edge types: prerequisite, reasoning-instance, abstraction, analogy. Production graph for agent physics reasoning.

Complete SI ↔ Gaussian unit conversion for electrodynamics. Replace ε₀, μ₀, c factors when translating between Jackson (SI), Landau (Gaussian), and other sources. Essential when using landau-graph (Gaussian) alongside electrodynamics skill (SI).

Universal methodology for distilling any mathematically rigorous discipline into a reasoning-constrained skill graph consumable by AI agents. Three-track extraction (reasoning mining, knowledge grounding, dependency mining), systematic rumination with blind-derivation validation, and cross-volume bridge construction. Complete worked example: Landau-Lifshitz 10-volume Course of Theoretical Physics (96 nodes, 188 edges).

Repository of mathematical theorems referenced by LandauGraph reasoning nodes. Stores complete proofs for results that are asserted but not derived in templates: Bertrand's theorem, Liouville-Arnold theorem, distribution identities, Airy asymptotics, and other mathematical infrastructure required for full derivations.

**You can learn how to set up cron here.** If the human asks you to do something on a schedule ("write a poem every hour," "remind me at 9am weekdays," "every 30 minutes check X"), this manual tells you how to wire it up using the host OS scheduler (macOS launchd / Linux systemd / crontab) — including the wake-by-mailbox-drop contract that lets a cron job invoke YOU, the hygiene rules that keep the script alive past its first failure, and a debugging walkthrough for when a scheduled job goes silent. Read this manual when: - The human asks for any time-driven recurring work and you don't yet know how to set up cron on this machine. **Read this BEFORE writing a plist or asking the human to do it manually.** - You are debugging a scheduled job that has gone silent, fires twice, kills its child process on exit, or otherwise misbehaves. - You are about to write the cron prompt body and want to know what belongs in the cron script versus what belongs in a custom skill the script invokes.

Operational guide for the `daemon` tool — how to inspect, debug, and reason about emanations (the agent fragments you spawn for parallel sub-tasks). Reach for this manual when: - You called `daemon(action="emanate", ...)` and the emanation has been running long enough that you want to know if it's stuck, slow, or working — without killing it on a hunch. - An emanation finished with `state=failed` or `state=timeout` and you need to figure out why. - You want to inspect on-disk artifacts of a past emanation (chat transcript, token spend, event log) — the folders persist forever and `daemon(action="list")` only shows currently-active runs. - You're trying to understand why your token totals don't match the sum of what your tools claim — daemon spend is tagged into your parent ledger and may explain the gap. This manual covers: the on-disk folder layout under `daemons/em-N-*/`, exactly which fields in `daemon.json` answer "is it stuck?" vs "is it thinking?", how to tail `chat_

Concise guide to the `knowledge` capability: private agent-owned memory in `<agent>/knowledge/<name>/KNOWLEDGE.md`, progressive disclosure through the prompt catalog, nested knowledge folders, and cross-references between entries. Read this when you need to create, organize, or load private knowledge, or when you need to explain how knowledge differs from portable skills.

Operational guide for the `mcp` capability — register, activate, update, deregister, and troubleshoot MCP (Model Context Protocol) servers in your agent. Single source of truth for both generic MCP setup AND the four kernel-curated LingTai addon MCPs (`imap`, `telegram`, `feishu`, `wechat`). Reach for this manual when: - The human asks to install, set up, configure, or remove an MCP server. Decision tree: is it kernel-curated (imap/telegram/feishu/wechat) → the `addons:` + `init.json mcp.<name>` workflow with the per-addon README is here. Is it third-party → the registry route OR the legacy `mcp/servers.json` route, both documented here. - The human asks to set up the `imap` / `telegram` / `feishu` / `wechat` addon, or any LingTai email/chat integration. **Step 1 is always**: fetch the addon's README with the bundled script — `~/.lingtai-tui/runtime/venv/bin/python3 .library/intrinsic/capabilities/mcp/scripts/find_readme.py <pkg-name>` (where `<pkg-name>` is `lingtai-imap`

Operational guide for the `skills` capability — your skill catalog's on-disk layout, how the `<available_skills>` XML in your system prompt is built, and the full authoring/publishing workflow for new skills. Reach for this manual when: - You're authoring a new skill in `.library/custom/<name>/` and need the frontmatter schema, the bundled template, the validator, or the "do create a skill / do NOT create a skill" decision rules. - You want to publish a custom skill to the network-shared library (`.library_shared/`) and need the `cp -r` recipe plus the admin stewardship norms. - A skill you expect to see isn't showing up in the catalog — the health-check workflow (`skills({"action": "info"})`) and the `intrinsic` vs `custom` directory split tell you what's wrong. - You want to pin a skill's body into your pad (so it survives the next molt and stays in the cached prefix) — the `psyche` pinning recipe lives here. - You're adding a new skills path source (e.g. a project-