Skill
adopt
Take a folder that already holds research — papers, notes, code, half-written drafts — into MAGI, without breaking what is already there.
@misaka16384View SkillData
Skill
/mentor
Drive a signed run while the person is away — choose the next step from the map, register it, have it done, record what came of it — so that any agent on any host can take over.
Install
$ npx -y skills add Misaka16384/Wikify --skill mentor --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
- You can call itInvoke it directly when you want it.
- Slash command
/mentor
Context preview
The summary Claude sees to decide when to auto-load this skill.
Drive a signed run while the person is away — choose the next step from the map, register it, have it done, record what came of it — so that any agent on any host can take over.
SKILL.md
mentor.SKILL.mdname: mentor
description: "Drive a signed run while the person is away — choose the next step from the map, register it, have it done, record what came of it — so that any agent on any host can take over."
commands:
mentor: "Carry a signed run forward, one registered step at a time."
origin: magi
mentor
When to use
`magi next` opens with `RUNNING`. You are the mentor whether or not you were in the discussion: the contract is frozen, and `magi run status <run>` is all of it.
Method
1. `magi run status <run>`. A step a dead session left open comes first: finish it, or `magi run result <run> <n> --abandoned --text "<why>"`. 2. Choose from the map — contract, questions, claims and their statuses — not from derivation details. Serve the contract's target; skip its "Not worth it". 3. Register before acting: `magi run step <run> --do "…" --because "…" --if-true "…" --if-false "…"`. Both outcomes lead to the same next move? That is trivia; do not take the step. 4. Have it done — a sub-agent where the host has them, yourself otherwise — through `research`, `compile`, `ingest`. Work is saved to `drafts/`, `tools/` and thread posts as it goes, never only in a context that can vanish. 5. `magi run result <run> <n> --text "…"`, linking what it produced. 6. A worker offers options and trying is cheap: take its favourite *and* the most ambitious. Every few steps pull back — the picture, the core insight (brute force said honestly), whether this is still promising. 7. Enough to write up, or only the report left (then the step is `--closing`): `magi run outline <run> --write`, and the `brief` skill takes it from there.
Rules
- **Never** edit the contract, sign, amend, bet for the person, close a line or
publish. A contract that looks wrong: post why, work another direction.
- **Never** ask the person anything mid-run; `magi next` parks what needs them.
Collect the questions workers could not ask and put them in the report.
- The CLI sets the concurrency (`max_parallel`): do not start what it refused.
- No state in the host's own task list, plan file or memory — another vendor's
agent cannot read them. **Partial** results are recorded as partial.
Read more
name: mentor description: "Drive a signed run while the person is away — choose the next step from the map, register it, have it done, record what came of it — so that any agent on any host can take over." commands: mentor: "Carry a signed run forward, one registered step at a time." origin: magi
mentor
When to use
`magi next` opens with `RUNNING`. You are the mentor whether or not you were in the discussion: the contract is frozen, and `magi run status <run>` is all of it.
Method
1. `magi run status <run>`. A step a dead session left open comes first: finish it, or `magi run result <run> <n> --abandoned --text "<why>"`. 2. Choose from the map — contract, questions, claims and their statuses — not from derivation details. Serve the contract's target; skip its "Not worth it". 3. Register before acting: `magi run step <run> --do "…" --because "…" --if-true "…" --if-false "…"`. Both outcomes lead to the same next move? That is trivia; do not take the step. 4. Have it done — a sub-agent where the host has them, yourself otherwise — through `research`, `compile`, `ingest`. Work is saved to `drafts/`, `tools/` and thread posts as it goes, never only in a context that can vanish. 5. `magi run result <run> <n> --text "…"`, linking what it produced. 6. A worker offers options and trying is cheap: take its favourite *and* the most ambitious. Every few steps pull back — the picture, the core insight (brute force said honestly), whether this is still promising. 7. Enough to write up, or only the report left (then the step is `--closing`): `magi run outline <run> --write`, and the `brief` skill takes it from there.
Rules
- **Never** edit the contract, sign, amend, bet for the person, close a line or
publish. A contract that looks wrong: post why, work another direction.
- **Never** ask the person anything mid-run; `magi next` parks what needs them.
Collect the questions workers could not ask and put them in the report.
- The CLI sets the concurrency (`max_parallel`): do not start what it refused.
- No state in the host's own task list, plan file or memory — another vendor's
agent cannot read them. **Partial** results are recorded as partial.
Ships withmagi
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
Get the whole plugin
Other skills on magi.
@misaka16384View SkillSkill
ask
Answer a question from this project's own knowledge — retrieved, read and cited, never from memory.
@misaka16384View SkillSkill
brief
Write the one thing a person reads after an unattended run — its report — and the lecture notes that prepare them for it, in words they already have.
@misaka16384View SkillSkill
compile
Compile raw sources into dense, interlinked reference and concept cards, and mine the concepts a compiled card left implicit.
@misaka16384View SkillSkill
discuss
Talk a run over with the person until it can be written down — motivation, directions, what is not worth it, when to stop, their taste — and they sign it.
@misaka16384View Skill