adopt
Take a folder that already holds research — papers, notes, code, half-written drafts — into MAGI, without breaking what is…
Entry point for a MAGI research project: run magi next, do what it says, and call the skill it names.
Claude Code
$ npx -y skills add misaka16384/magi --skill magi --agent claude-code
Codex
$ npx skills add misaka16384/magi --skill magi -a codex -y
Cursor
$ npx skills add misaka16384/magi --skill magi -a cursor -y
Gemini CLI
$ npx skills add misaka16384/magi --skill magi -a gemini-cli -y
GitHub Copilot
$ npx skills add misaka16384/magi --skill magi -a github-copilot -y
OpenCode
$ npx skills add misaka16384/magi --skill magi -a opencode -y
Windsurf
$ npx skills add misaka16384/magi --skill magi -a windsurf -y
name: magi description: "Entry point for a MAGI research project: run magi next, do what it says, and call the skill it names." commands: magi: "Work out what this research project needs next, and do it." origin: magi
Any turn in a MAGI project where you do not already know what you are doing.
1. `magi next` — it reads the notes and ranks what is owed. It never acts. 2. Do the first item, unless the human just asked for something else: their sentence outranks the list. Everything under **"For the person"** goes to them in **one message**, and **asking for a prediction means showing the claim** — `next` prints it under the item, `--json` has it in `bets_waiting`. A slug is not a question anybody can answer. 3. When an item names a skill, call it. When it names a command, run it. 4. A result that came before its note is a finding: open it with `--found`, or mark an existing one `magi thread found <slug>`. Neither is asked to bet. A reviewer's `restate` is yours: fix the words as its post says, then `magi thread status <slug> supported --text '<what changed>'`. 5. Anything the human says in passing worth keeping goes into `inbox/notes.md`, one line, untidied; a later `magi next` files it. 6. Before stopping: `magi sync --close`. It refuses while something happened that nobody wrote down — clear each with `magi thread post <slug> --text '<what happened>' --host human` or `magi decide --about <slug> --text '<their words>'`, then close again.
from the notes; a step you made up is derived from nothing.
<status> --text '<why>'` writes the flip and its reason together, or neither.
`magi close <line>`, or `magi decide --about <slug> --text '<their words>'`. **Never** redo such a flip unsigned to pass the close gate; each unsigned repeat is one more signature owed.
are not advice. Stuck: `magi guide --search "<the error>"`, never a guessed flag.
MAGI 是一个 agent-native 的科研工作环境:人是驾驶员,LLM agent 是机体,确定性的 magi CLI 是拘束具——三者同步率越高,科研越快。它把学术论文(PDF/LaTeX)摄入、编译为 Obsidian 兼容的概念卡片项目,并用三核架构管理完整的科研状态: 进入任意项目先跑 magi sync——它输出同步率、三核状态和逐条可执行的修复提示。magi radar 是文献雷达:定时发现相关新论文,并侦察"知识上应引用我方论文却未引用"的候选。同一份 skills 通吃 Claude
Take a folder that already holds research — papers, notes, code, half-written drafts — into MAGI, without breaking what is…
Answer a question from this project's own knowledge — retrieved, read and cited, never from memory.
Write the one thing a person reads after an unattended run — its report — and the lecture notes that prepare them for it, in…
Compile raw sources into dense, interlinked reference and concept cards, and mine the concepts a compiled card left implicit.
Talk a run over with the person until it can be written down — motivation, directions, what is not worth it, when to stop, their…
Turn whatever the human has — a PDF, a link, a DOI, a screenshot of a citation — into raw/ sources this project can compile.
Use this skill to extract structured Markdown/JSON from PDFs and document images—tables with cell-level precision, formulas as…
Estimates intracellular metabolic fluxes from steady-state carbon-13 isotope-tracing measurements using validated atom maps,…