Connect Agent Skills & MCP

Agent Skills • MCP • OpenAPI 3.1

⚡

Integrate with AI Coding Agents

Protect your agents from repeating known LLM failure modes and anti-patterns

Prompt Confessional exports full real-time machine specifications across Agent Skills, Model Context Protocol (MCP), OpenAPI 3.1, and LLMs.txt.

🚀1-Click Integration Recipes

Pick your agent or IDE

Universal Skills CLI (skills.sh)Recommended

Install the public aifails skill across Claude Code, Pi/OMP, Cursor, and Codex with one command:

npx skills add sirmews/aifails

Or run in one-shot mode:

npx skills use sirmews/aifails@aifails

Anthropic Claude Desktopclaude_desktop_config.json

Add this block to your claude_desktop_config.json under mcpServers:

{
  "mcpServers": {
    "aifails": {
      "url": "https://aifails.wtf/mcp"
    }
  }
}

Cursor & Windsurf.cursor/mcp.json

In Cursor, go to Settings → Features → MCP → Add New MCP Server with Type HTTP, or add to .cursor/mcp.json:

{
  "mcpServers": {
    "aifails": {
      "url": "https://aifails.wtf/mcp"
    }
  }
}

OpenCodeopencode.json

Add this block to your project's opencode.json or global config:

{
  "mcp": {
    "servers": {
      "aifails": {
        "type": "http",
        "url": "https://aifails.wtf/mcp"
      }
    }
  }
}

Included Capabilities & Tools

get_anti_patterns

Queries known LLM traps by framework (e.g. Tailwind, Rust, D1, Next.js) and returns actionable negative prompt rules.

get_random_fail

Fetches a random prompt failure with developer reaction and community prompt fixes in clean Markdown.

submit_fail

Anonymously contributes prompt failures directly from agent sessions with automated Gitleaks secret scrubbing.

Machine Specifications & Direct Descriptors

📄 OpenAPI 3.1 (JSON)/openapi.json ⚡ Skills Index (RFC v0.2.0)/.well-known/agent-skills/index.json 📦 Raw Skill Definition/skill.md 🌐 RFC 9727 API Catalog/.well-known/api-catalog 🤖 Full Anti-Pattern Catalog/llms-full.txt