Claude-skill-registry-data minimax-mcp-usage
Optimal patterns for MiniMax MCP tools (web_search + understand_image)
install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry-data
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry-data "$T" && mkdir -p ~/.claude/skills && cp -r "$T/data/minimax-mcp-usage" ~/.claude/skills/majiayu000-claude-skill-registry-data-minimax-mcp-usage && rm -rf "$T"
manifest:
data/minimax-mcp-usage/SKILL.mdsource content
ultrathink - Take a deep breath. We're not here to write code. We're here to make a dent in the universe.
The Vision
MiniMax queries should be lean, accurate, and decisive.
Your Work, Step by Step
- Choose tool: web_search vs understand_image.
- Craft query: Specific, time-bound, and focused.
- Run analysis: Collect results with minimal noise.
- Synthesize: Convert outputs into clear guidance.
Ultrathink Principles in Practice
- Think Different: Use the lowest-cost path to truth.
- Obsess Over Details: Precision in queries matters.
- Plan Like Da Vinci: Decide intent before search.
- Craft, Don't Code: Keep results actionable.
- Iterate Relentlessly: Refine queries until clear.
- Simplify Ruthlessly: Cut vague or broad requests.
MiniMax MCP Usage Patterns (v2.24)
This skill documents optimal usage patterns for MiniMax MCP tools.
Available Tools
1. mcp__MiniMax__web_search
Purpose: Web search with 8% cost of alternatives
Input:
query: string # 3-5 keywords, include year for recent topics
Output:
{ "organic": [{ "title", "link", "snippet", "date" }], "related_searches": [{ "query" }] }
Optimal Patterns:
# Good: Specific, time-bounded mcp__MiniMax__web_search: query: "React 19 useOptimistic hook examples 2025" # Good: Error-focused mcp__MiniMax__web_search: query: "TypeError cannot read property undefined Next.js" # Bad: Too vague mcp__MiniMax__web_search: query: "javascript" # Too broad
2. mcp__MiniMax__understand_image
Purpose: Image analysis for debugging and review
Input:
prompt: string # Clear, specific question about the image image_source: string # Local path (no @) or HTTPS URL
Optimal Patterns:
# Good: Specific analysis request mcp__MiniMax__understand_image: prompt: "Identify the exact error message and stack trace in this screenshot" image_source: "/tmp/error.png" # Good: UI review mcp__MiniMax__understand_image: prompt: "List all accessibility violations in this form design" image_source: "./mockup.png" # Bad: Vague prompt mcp__MiniMax__understand_image: prompt: "What's this?" # Too vague image_source: "./image.png"
Integration with Ralph Loop
# Research phase: Use web_search Task: prompt: | Research latest patterns for $TOPIC using mcp__MiniMax__web_search. Compile findings into structured report. # Debugging phase: Use understand_image Task: prompt: | Analyze error screenshot at $PATH using mcp__MiniMax__understand_image. Identify root cause and suggest fixes.
Cost Analysis
| Operation | MiniMax MCP | Gemini CLI | Savings |
|---|---|---|---|
| Web search | ~$0.008 | ~$0.06 | 87% |
| Image analysis | ~$0.01 | N/A | New capability |
When NOT to Use
| Scenario | Alternative |
|---|---|
| US-only search | WebSearch (free) |
| Code search | ast-grep MCP (v2.23) |
| Long-form generation | Gemini CLI (1M context) |
| Real-time data | Native WebFetch |
Troubleshooting
| Issue | Solution |
|---|---|
| "API key invalid" | Check MINIMAX_API_KEY in ~/.claude.json |
| "Image too large" | Compress to <20MB |
| "Format not supported" | Convert to JPEG/PNG/WebP |
| "No results" | Refine query with more keywords |