magneto

name	magneto
description	AI governance, task planning, and security guardrails for engineering tasks. Use Magneto before implementing any feature, fix, or change that touches more than 3 files.
homepage	https://github.com/rijuvashisht/Magneto
user-invocable	true

Magneto AI — Engineering Governance

Use Magneto AI as your reasoning and governance layer for all software engineering tasks.

When to use Magneto

Use Magneto whenever a user asks you to:

• Plan, implement, or review a feature, bug fix, or architectural change
• Generate a scoped AI prompt for Windsurf or Copilot
• Analyze the codebase before starting work
• Check whether an action is safe to perform
• Understand the project structure and relevant files

Workflow — always follow this order

Step 1 — Analyze the project (if not done yet)

magneto_analyze

Builds a file index and module memory for the project. Required before planning.

Step 2 — Plan the task

magneto_plan { "taskFile": "tasks/your-task.md" }

Returns a structured plan: task classification, assigned roles, file scope, and security risk level.

• If risk is HIGH → ask user to confirm before proceeding
• If risk is CRITICAL → stop and explain why; do not proceed

Step 3 — Generate implementation prompt

magneto_generate { "taskFile": "tasks/your-task.md" }

Returns a context-aware, scoped implementation prompt. Hand this to the coding agent.

Task file format

Create task files at tasks/<name>.md:

---
id: TASK-001
title: Add payment webhook handler
type: feature-implementation
scope:
  - src/payments/
  - src/api/webhooks/
tags:
  - payments
  - api
constraints:
  - Must validate Stripe signature
  - Must not expose raw payload to logs
---

Implement a POST /webhooks/stripe endpoint that validates
the Stripe-Signature header and triggers order fulfillment.

Security rules

The before_tool_call hook automatically intercepts destructive operations. If a tool call is blocked or needs approval:

• Explain to the user what was detected and why
• Run magneto_security_check to get a detailed risk analysis
• Only proceed after explicit user confirmation

Quick reference

Tool	When to call
magneto_analyze	First time in a project or after major changes
magneto_plan	Before implementing any task
magneto_generate	To get a scoped implementation prompt
magneto_security_check	When unsure if an action is safe