magi

Original：🇺🇸 English

Translated

Use only when the user explicitly requests brainstorming, evaluating architecture choices, or comparing options where no single concern dominates

2installs

Sourcevinta/hal-9000

Added on2026-02-25

NPX Install

npx skill4agent add vinta/hal-9000 magi

SKILL.md Content

View Translation Comparison →

MAGI

Three-agent deliberation system. Spawns Scientist, Mother, and Woman teammates with competing value lenses to explore a question in parallel, then consolidates their proposals for the user.

Checklist

Follow these steps in order.

Clarify the question
Set up agent team
Parallel exploration
Consolidate and present options
Wait for user decision (implement, debate, or done)
Tear down agent team

Workflow

dot

digraph magi {
    rankdir=TB;
    node [shape=box, style=rounded];

    clarify [label="Lead clarifies question\n(AskUserQuestion if underspecified)"];
    setup [label="Setup 3-agent team"];
    explore [label="Teammates explore in parallel\n(read project, search online, propose 2-3 approaches)"];
    consolidate [label="Lead consolidates proposals\n+ lightweight stance per teammate"];
    present [label="Present options to user\n(AskUserQuestion: Implement / Debate / Done)"];
    decide [label="User decision" shape=diamond];
    debate [label="Debate round\n(peer-to-peer, 1 exchange)"];
    teardown [label="Shut down agent team"];
    handoff [label="Handoff to writing-plans"];

    clarify -> setup;
    setup -> explore;
    explore -> consolidate;
    consolidate -> present;
    present -> decide;
    decide -> debate [label="debate"];
    decide -> teardown [label="done"];
    decide -> teardown [label="implement"];
    teardown -> handoff [label="implement"];
    debate -> present;
}

The Process

User Question

$ARGUMENTS

Clarify

Before spawning teammates, the lead asks via

AskUserQuestion

to understand the idea or topic:

Ask questions one at a time to refine the idea
Prefer multiple choice questions when possible, but open-ended is fine too
Only one question per message - if a topic needs more exploration, break it into multiple questions
Focus on understanding: purpose, constraints, success criteria

Skip if the question is already clear and actionable. Include all clarified context in teammate spawn prompts.

Setup

Read all 3 template files: MAGI-1.md, MAGI-2.md, MAGI-3.md

TeamCreate

with a descriptive team name

magi-{topic}

(e.g.,

magi-auth-strategy

)

Spawn all 3 teammates in a single message (3 parallel
```
Task
```
calls):
- ```
subagent_type
```
  :
```
general-purpose
```
- ```
team_name
```
  : team name from above
- ```
name
```
  :
```
scientist
```
  /
```
mother
```
  /
```
woman
```
- ```
prompt
```
  : template content + user's question + clarified context from Clarify step + Teammate Checklist (below)

Teammates don't inherit the lead's conversation history -- include all context in the spawn prompt.

Explore in Parallel

Teammates begin working immediately upon spawning. The lead's role is coordination only:

Wait for teammates to send their proposals via
```
SendMessage
```
If a teammate sends a clarifying question, forward it to the user via
```
AskUserQuestion
```
-- note which teammate asked. Do NOT answer teammate questions yourself; only the user answers.
Do NOT explore, research, or generate proposals yourself

Teammate Checklist

Each teammate MUST create a task for each step and complete them in order:

Explore project state -- check files, docs, recent commits
Ask clarifying questions -- if anything is unclear, ask the lead to ask the user. The lead should note which teammate asked. DO NOT skip this.
Search online -- find relevant prior art, docs, discussions. DO NOT skip this.
Evaluate/generate options -- if user is open-ended, generate from scratch; if user supplies options, evaluate those AND propose alternatives. Surface non-obvious ideas -- discover what's missing, don't just analyze what's given.
Propose 2-3 approaches -- with trade-offs from your persona's lens
Tag top pick -- one-line rationale for your recommended option
Report to lead -- send your proposals and top pick to the lead via
```
SendMessage
```

Consolidate + Present

Lead collects all proposals from the 3 teammates, then:

Deduplicates similar proposals (attributing to all teammates who proposed it)
Groups by theme if there are many proposals
Presents each option with:
- Which teammate(s) proposed it
- Trade-off analysis from each perspective
- Who tagged it as their top pick and why
Asks the user via
```
AskUserQuestion
```
to select an option (one option per choice)
Asks the user via
```
AskUserQuestion
```
what to do next:
- Implement — triggers teardown + handoff to
```
writing-plans
```
  with the selected option
- Debate — teammates critique each other's proposals (triggers debate round below)
- Done — shut down the agent team, no further action

Optional Debate (user-triggered)

Only runs if the user requests it. When triggered:

Send the consolidated option list to all teammates
Each teammate sends direct messages to each other teammate critiquing their proposals
Each teammate gets one response to defend or concede
Lead collects updated stances and re-presents

Teardown

When to tear down

User selects Implement
User selects Done

When NOT to tear down

After presenting proposals -- the debate loop requires live teammates

Shutdown sequence

```
shutdown_request
```
to each teammate (Scientist, Mother, Woman)
Wait for all shutdown approvals
```
TeamDelete
```
(fails if teammates are still active)

Handoff (implement path only)

After teardown, invoke

writing-plans

skill with the chosen option(s) as context.