adopt
Take a folder that already holds research — papers, notes, code, half-written drafts — into MAGI, without breaking what is…
Compile raw sources into dense, interlinked reference and concept cards, and mine the concepts a compiled card left implicit.
Claude Code
$ npx -y skills add misaka16384/magi --skill compile --agent claude-code
Codex
$ npx skills add misaka16384/magi --skill compile -a codex -y
Cursor
$ npx skills add misaka16384/magi --skill compile -a cursor -y
Gemini CLI
$ npx skills add misaka16384/magi --skill compile -a gemini-cli -y
GitHub Copilot
$ npx skills add misaka16384/magi --skill compile -a github-copilot -y
OpenCode
$ npx skills add misaka16384/magi --skill compile -a opencode -y
Windsurf
$ npx skills add misaka16384/magi --skill compile -a windsurf -y
name: compile description: "Compile raw sources into dense, interlinked reference and concept cards, and mine the concepts a compiled card left implicit." commands: compile: "Compile raw sources into wiki cards, and enrich thin ones." origin: magi
`magi next` reports a compile backlog, or `magi wiki uncompiled` lists sources with no card.
1. `magi wiki uncompiled` — the backlog. Nothing else decides what is next. 2. One sub-agent per source, at most 10 at once. Each reads the raw file and writes one card using `<SKILL_DIR>/templates/paper_template.md`, creating concepts with `magi wiki add-concept --name/--source/--content` as it goes. 3. Every claim's `SOURCE:` points at the `raw/` file, never at another card. 4. Thin cards: `magi wiki placeholders <card.md>` and `magi stats concept-density <card.md>`, one card at a time. Under ~5 links or ~2 per 1000 words, mine the source for the concepts it actually contains — never invent one it does not. 5. `magi lint --fix` first, then `magi wiki reindex`. That order matters: reindex reads what lint has already normalised. 6. `magi graph build` and `magi index` to make it findable.
sub-agent calls" is a sentence a person can stop. Ten concurrent is the ceiling.
Awaiting synthesis]` without saying so in your report. **Partial** work is reported as partial.
them to the human once, when the batch reports back.
recompile.
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…
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.
Entry point for a MAGI research project: run magi next, do what it says, and call the skill it names.
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,…