magnific-mcp skill for Claude Code

A best-practices guide that teaches Claude how to use the Magnific MCP server effectively — covering model selection, credit budgeting, character consistency via the Library, image→video chaining, and approval rules for expensive operations.

This is a Skill, not an MCP. It sits on top of the existing Magnific MCP (which Magnific provides) and gives Claude the meta-knowledge to drive it well.

Prerequisites

• An active Magnific subscription.
• The Magnific MCP connected to your Claude client (Claude Code, Claude Desktop, etc.). The skill auto-detects whether the MCP is registered as magnific, magnific-personal, or magnific-work.

Install

Clone into your Claude skills directory:

# macOS / Linux
git clone https://github.com/bnayoo3-cyber/magnific-mcp-skill ~/.claude/skills/magnific-mcp

# Windows (PowerShell)
git clone https://github.com/bnayoo3-cyber/magnific-mcp-skill $env:USERPROFILE\.claude\skills\magnific-mcp

Restart your Claude client. The skill becomes active automatically — Claude will load it whenever you mention Magnific, generation, credits, or related tasks.

First-run setup

The first time you ask Claude to generate something via Magnific, it will ask three quick questions:

1. Usage alerts — warn at a credit threshold? (none / percent / absolute)
2. Credit tracking — auto-update a local budget.json after each call?
3. Cost preference — quality-first, balanced, or careful?

Your answers persist in ~/.claude/skills/magnific-mcp/user_config.json and shape every subsequent decision.

What's inside

File	Purpose
SKILL.md	The guide itself — 10 sections of accumulated best practice.
user_config.example.json	Template for your preferences.
budget.example.json	Template for credit tracking.

Customising

SKILL.md is plain markdown — edit it freely. If you discover a better default through use, propose it via PR. Keep changes provider-neutral (no account-tier specifics; those belong in your local budget.json).

License

MIT.