Feed Type: `prompt.llmfeed.json`

Purpose

This feed encapsulates a structured prompt that can be shared, interpreted, replayed or executed by an LLM or agent.

The Evolution of Prompt Engineering

We acknowledge that Markdown-structured prompts have become the mainstream approach for power users and prompt engineers. Moving from basic plaintext to Markdown with clear sections, variables, and formatting instructions was already a significant improvement:

# Customer Analysis Prompt

## Context
You are analyzing digital marketing data for a B2B SaaS company.

## Instructions
1. Compare performance vs objectives
2. Identify top-performing channels
3. Recommend budget reallocations

## Output Format
- Executive Summary (3 bullets)
- Detailed Analysis (structured sections)
- Action Items (prioritized list)

This Markdown approach has rightfully become the standard because it provides:

✅ Visual structure for human readability
✅ Sections and hierarchy for logical organization
✅ Variables and placeholders for reusability
✅ Formatting hints for output control

Why JSON Goes Further

However, JSON-structured prompts (prompt.llmfeed.json) take this evolution to its logical conclusion by adding:

🔐 Cryptographic signatures - Ensuring integrity and authorship
📜 Certification - Third-party validation for compliance
🔌 Direct API execution - No copy-paste, automatic orchestration
💰 Cost tracking - Built-in economics and billing metadata
📊 Performance metrics - Measurable success rates and optimization
🔄 Workflow integration - Native support for N8N, Zapier, Make
📝 Session generation - Automatic audit trails for every execution

While Markdown prompts are excellent for human collaboration, JSON prompts enable machine orchestration at scale. It's not about replacing Markdown prompts—it's about making them executable, verifiable, and economically trackable in the agentic economy.

Typical Use Cases

Prompt engineering documentation
Invocation of external services or agent actions
Instructing LLMs to generate other .llmfeed.json types
Sharing reproducible queries across agents or tools
Professional prompt libraries and marketplaces
Automated workflow execution via N8N, Zapier, Make
API-driven prompt orchestration
Cost-tracked enterprise deployments

Canonical Structure

{
  "feed_type": "prompt",
  "metadata": {
    "title": "Generate a session feed",
    "origin": "https://tool.llmfeed.org",
    "author": "Alex Chen",
    "created_at": "2025-06-10T14:30:00Z"
  },
  "intent": "export current session as JSON",
  "context": "User is finishing a chat and wants to save the reasoning path.",
  "precision_level": "ultra-strict",
  "result_expected": "session",
  "process_mode": "prepare-for-another",
  "prompt_body": "You are an LLM that supports LLMFeed. Please generate a session feed with context, output and decisions.",
  "trust": {
    "signed_blocks": ["metadata", "prompt_body", "trust"],
    "scope": "public",
    "certifier": "https://llmca.org"
  },
  "signature": {
    "value": "abc123...",
    "created_at": "2025-06-10T14:30:00Z"
  }
}

Core Fields

Field	Required	Description
`prompt_body`	✅	The actual instruction to the LLM
`intent`	✅	What the user or system expects
`context`	⚠️	Extra info the LLM should consider
`precision_level`	⚠️	`"raw"`, `"strict"`, `"ultra-strict"`
`process_mode`	⚠️	`"instruct"`, `"fill-and-execute"`, `"prepare-for-another"`
`result_expected`	⚠️	`"text"`, `"feed"`, `"code"`, `"session"`
`attachments[]`	⚠️	Optional examples, templates, context
`audience`	⚠️	If only for LLM, wrapper, user etc.

🔌 API Execution & Orchestration

From Specification to Execution

The true power of prompt.llmfeed.json emerges when combined with API execution platforms like N8N, Zapier, or Make. These prompts aren't just documentation—they're executable specifications.

graph LR
    A[prompt.llmfeed.json] -->|Parse & Validate| B[Orchestration Platform]
    B -->|Extract prompt_body| C[LLM API]
    C -->|Execute with context| D[Response]
    D -->|Format as| E[session.llmfeed.json]
    
    B -->|Verify| F[Signature]
    B -->|Check| G[Certification]
    
    style A fill:#f9f,stroke:#333,stroke-width:2px
    style E fill:#9f9,stroke:#333,stroke-width:2px

API Integration Pattern

{
  "feed_type": "prompt",
  "metadata": {
    "title": "Customer Support Analyzer",
    "api_ready": true
  },
  "execution_hints": {
    "preferred_apis": ["anthropic", "openai", "google"],
    "api_parameters": {
      "anthropic": {
        "model": "claude-3-5-sonnet-20241022",
        "max_tokens": 2000,
        "temperature": 0.7
      },
      "openai": {
        "model": "gpt-4-turbo",
        "max_tokens": 2000,
        "temperature": 0.7
      }
    },
    "timeout_ms": 30000,
    "retry_policy": {
      "max_attempts": 3,
      "backoff_ms": [1000, 2000, 4000]
    }
  },
  "prompt_body": "Analyze this customer support ticket...",
  "webhook_endpoints": {
    "on_success": "https://api.example.com/webhook/success",
    "on_failure": "https://api.example.com/webhook/failure",
    "on_certification_expired": "https://api.example.com/webhook/cert-expired"
  }
}

Direct API Execution

Platforms can execute prompt.llmfeed.json directly via API calls:

// N8N/Node.js Example
async function executeLLMFeedPrompt(feedUrl) {
  // 1. Fetch the prompt feed
  const feed = await fetch(feedUrl).then(r => r.json());
  
  // 2. Verify signature & certification
  if (!await verifyTrust(feed)) {
    throw new Error('Trust verification failed');
  }
  
  // 3. Extract execution parameters
  const apiConfig = feed.execution_hints?.api_parameters?.anthropic || {};
  
  // 4. Execute via API
  const response = await fetch('https://api.anthropic.com/v1/messages', {
    method: 'POST',
    headers: {
      'x-api-key': process.env.ANTHROPIC_KEY,
      'anthropic-version': '2023-06-01',
      'content-type': 'application/json'
    },
    body: JSON.stringify({
      model: apiConfig.model || 'claude-3-5-sonnet-20241022',
      max_tokens: apiConfig.max_tokens || 2000,
      temperature: apiConfig.temperature || 0.7,
      system: feed.context || '',
      messages: [{
        role: 'user',
        content: feed.prompt_body
      }]
    })
  });
  
  // 5. Return as session.llmfeed.json
  return formatAsSession(feed, response);
}

API Cost Tracking

"api_economics": {
  "estimated_tokens": {
    "input": 500,
    "output": 1500,
    "total": 2000
  },
  "estimated_cost": {
    "anthropic": "$0.024",
    "openai": "$0.030",
    "google": "$0.020"
  },
  "billing_metadata": {
    "project_id": "analytics-automation",
    "cost_center": "operations",
    "invoice_reference": "prompt-001"
  }
}

Batch Processing Support

"batch_execution": {
  "supported": true,
  "max_batch_size": 100,
  "parallelization": {
    "max_concurrent": 5,
    "rate_limit_per_minute": 60
  },
  "aggregation_mode": "individual_sessions",
  "batch_webhook": "https://api.example.com/batch/complete"
}

API Authentication Patterns

"api_authentication": {
  "method": "delegated",
  "credential_references": {
    "anthropic": "vault://credentials/anthropic-api-key",
    "openai": "env://OPENAI_API_KEY",
    "custom": "https://auth.example.com/token"
  },
  "oauth_flow": {
    "enabled": false,
    "provider": "https://auth.provider.com",
    "scopes": ["read:prompts", "execute:prompts"]
  }
}

Execution Monitoring

"monitoring": {
  "metrics_endpoint": "https://metrics.example.com/prompts",
  "tracked_metrics": [
    "execution_time_ms",
    "token_usage",
    "success_rate",
    "trust_score",
    "api_costs"
  ],
  "alerting": {
    "threshold_exceeded": "webhook://alerts/threshold",
    "certification_expiring": "webhook://alerts/cert-expiry",
    "execution_failed": "webhook://alerts/failure"
  }
}

🔄 Platform Integration Examples

N8N Workflow Integration

{
  "feed_type": "prompt",
  "metadata": {
    "title": "N8N-Optimized Analytics Processor",
    "platform": "n8n"
  },
  "n8n_integration": {
    "workflow_template": "https://n8n.io/workflows/analytics-template",
    "required_nodes": ["http-request", "code", "anthropic"],
    "environment_variables": [
      "ANTHROPIC_API_KEY",
      "DATA_SOURCE_ID"
    ],
    "trigger_type": "webhook",
    "webhook_path": "/execute/analytics"
  }
}

Zapier Integration

{
  "zapier_integration": {
    "zap_template": "https://zapier.com/shared/analytics-template",
    "trigger": "new_row_in_sheets",
    "actions": ["parse_llmfeed", "call_anthropic", "update_sheet"]
  }
}

Make (Integromat) Integration

{
  "make_integration": {
    "scenario_blueprint": "https://make.com/blueprints/analytics-001",
    "modules": ["json", "http", "claude", "sheets"],
    "data_structure": "llmfeed_prompt_v1"
  }
}

🚀 Real-World API Execution Pipeline

Complete Execution Flow

sequenceDiagram
    participant User
    participant Platform as N8N/Zapier
    participant Validator as Trust Validator
    participant API as LLM API
    participant Storage as Data Storage
    
    User->>Platform: Submit prompt.llmfeed.json
    Platform->>Validator: Verify signature & cert
    Validator-->>Platform: Trust score: 0.95
    
    Platform->>API: Execute prompt with context
    API-->>Platform: Response + tokens used
    
    Platform->>Storage: Save session.llmfeed.json
    Platform-->>User: Return formatted results
    
    Platform->>Platform: Track costs & metrics
    Platform->>Platform: Check webhooks

Healthcare-Specific API Considerations

For sensitive domains requiring regulatory compliance:

"healthcare_api_compliance": {
  "data_residency": "eu-west-1",
  "encryption": "aes-256-gcm",
  "audit_logging": true,
  "gdpr_compliant": true,
  "anonymization": {
    "pii_removal": true,
    "patient_data_handling": "pseudonymized"
  },
  "api_endpoints": {
    "primary": "https://eu.api.anthropic.com",
    "fallback": "https://eu2.api.anthropic.com",
    "never_use": ["us-east-1", "asia-pacific"]
  }
}

📊 Performance & Optimization

API Response Caching

"caching_policy": {
  "enabled": true,
  "ttl_seconds": 3600,
  "cache_key_components": ["prompt_body_hash", "context_hash", "api_params"],
  "invalidation_triggers": ["certification_renewal", "prompt_update"],
  "cache_storage": "redis://cache.example.com"
}

Load Balancing Across APIs

"load_balancing": {
  "strategy": "weighted_round_robin",
  "api_weights": {
    "anthropic": 0.5,
    "openai": 0.3,
    "google": 0.2
  },
  "failover_order": ["anthropic", "openai", "google"],
  "health_check_endpoint": "/health"
}

🔐 Security in API Execution

API Key Management

"security": {
  "api_key_rotation": {
    "frequency_days": 90,
    "rotation_webhook": "https://security.example.com/rotate",
    "grace_period_hours": 24
  },
  "key_storage": {
    "method": "vault",
    "provider": "hashicorp",
    "path": "/secret/llmfeed/api-keys"
  },
  "rate_limiting": {
    "per_minute": 60,
    "per_hour": 1000,
    "per_day": 10000
  }
}

Trust & Ownership

Basic Signing

"trust": {
  "signed_blocks": ["metadata", "prompt_body", "trust"],
  "scope": "public"
},
"signature": {
  "value": "signature_hash_here",
  "created_at": "2025-06-10T14:30:00Z"
}

Certification (Optional)

"certification": {
  "issuer": "https://llmca.org",
  "cert_id": "llmca-prompt-2025-001",
  "certified_blocks": ["prompt_body", "performance_metrics"],
  "issued_at": "2025-06-10T10:00:00Z",
  "expires_at": "2026-06-10T10:00:00Z"
}

Why sign prompts?

✅ Prove authorship and prevent tampering
✅ Build reputation as a prompt engineer
✅ Enable marketplaces for certified prompts
✅ Establish trust for sensitive use cases
✅ Enable automated execution with confidence

Agent Behavior

An agent that receives this feed should:

Parse the prompt_body and execute it
Respect precision_level and process_mode
Attach any referenced templates or context
Return a structured response as declared in result_expected
Verify signatures if trust is required
Track execution costs and generate session feeds

Simple Examples

Educational Prompt

{
  "feed_type": "prompt",
  "metadata": {
    "title": "Python Code Explainer",
    "author": "CS101 Team"
  },
  "intent": "explain Python code for beginners",
  "prompt_body": "Explain this Python code step-by-step for a beginner: [CODE]",
  "audience": ["student", "llm"],
  "result_expected": "text"
}

API Documentation Generator

{
  "feed_type": "prompt",
  "metadata": {
    "title": "API Doc Generator",
    "origin": "https://devtools.example.com"
  },
  "intent": "generate API documentation",
  "prompt_body": "Generate clear API documentation for this endpoint: [ENDPOINT_DATA]",
  "result_expected": "markdown",
  "attachments": [
    {
      "name": "doc_template.md",
      "description": "Standard documentation template"
    }
  ]
}

📚 Advanced Features

Professional Use Cases

Medical Consultation Assistant

{
  "feed_type": "prompt",
  "metadata": {
    "title": "Emergency Triage Assistant",
    "author": "Dr. Sarah Chen, MD",
    "institution": "Regional Medical Center"
  },
  "professional_validation": {
    "medical_board_approved": true,
    "peer_reviewed": true,
    "liability_coverage": "institutional_malpractice_policy"
  },
  "prompt_body": "Assess patient symptoms and provide triage recommendations following emergency medicine protocols...",
  "usage_restrictions": {
    "requires_medical_license": true,
    "human_oversight_required": true,
    "emergency_use_only": false
  },
  "api_economics": {
    "estimated_cost": "$0.15",
    "billing_code": "TRIAGE_ASSIST_001"
  }
}

Legal Document Analysis

{
  "feed_type": "prompt",
  "metadata": {
    "title": "Contract Risk Analyzer",
    "author": "LegalTech Solutions Inc."
  },
  "commercial_licensing": {
    "license_type": "professional",
    "pricing": "$0.10_per_analysis",
    "client_restrictions": "law_firms_only"
  },
  "prompt_body": "Analyze this contract for potential risks and compliance issues...",
  "execution_hints": {
    "preferred_apis": ["anthropic"],
    "api_parameters": {
      "anthropic": {
        "model": "claude-3-5-sonnet-20241022",
        "temperature": 0.2
      }
    }
  }
}

Intellectual Property & Licensing

Copyright Protection

"intellectual_property": {
  "copyright": "© 2025 Prompt Engineer Name",
  "license": "Creative Commons Attribution 4.0",
  "attribution_required": true,
  "commercial_use": "permitted_with_license"
}

Commercial Licensing

"commercial_licensing": {
  "available": true,
  "pricing_model": "per_use",
  "base_rate": "$0.01_per_invocation",
  "volume_discounts": [
    {"min_uses": 1000, "discount": "10%"},
    {"min_uses": 10000, "discount": "25%"}
  ],
  "contact": "licensing@promptcompany.com"
}

Performance & Analytics

Performance Metrics

"performance_metrics": {
  "accuracy_benchmark": "94.7%",
  "average_response_time": "2.3s",
  "user_satisfaction": "4.6/5.0",
  "total_uses": 15847,
  "success_rate": "91.2%",
  "api_cost_average": "$0.023"
}

LLM Compatibility

"llm_compatibility": {
  "gpt_4": {"score": 0.95, "optimal_temp": 0.3},
  "claude_3": {"score": 0.92, "optimal_temp": 0.2},
  "gemini": {"score": 0.89, "requires_adaptation": true}
}

Behavioral Controls

Agent Guidance

"agent_guidance": {
  "interaction_tone": "professional",
  "consent_hint": "Ask user before accessing sensitive data",
  "risk_tolerance": "low",
  "fallback_behavior": "escalate_to_human"
}

Trigger Targets

"trigger_targets": [
  {
    "type": "llmfeed",
    "feed_type": "export", 
    "url": "https://example.org/exports/results.llmfeed.json"
  },
  {
    "type": "action",
    "label": "Generate Report",
    "href": "/tools/report-generator"
  }
]

Best Practices

For Basic Users

Start simple - just prompt_body and intent
Add trust block for any shared prompts
Use clear metadata for discovery
Test with target LLMs before publishing
Add execution_hints for API optimization

For Professional Use

Always sign commercially-used prompts
Seek certification for high-risk domains
Include performance metrics for transparency
Follow licensing requirements in your jurisdiction
Test cross-platform compatibility
Implement cost tracking from day one
Generate session feeds for all executions

For Marketplaces

Implement proper attribution systems
Verify signatures before listing
Respect usage restrictions and licensing terms
Provide performance analytics to buyers
Support API execution metrics

For Automation Platforms

Cache signature verifications (TTL: 1 hour)
Implement retry logic with exponential backoff
Track token usage and costs per execution
Generate session.llmfeed.json for audit trails
Support multiple API providers for redundancy

MIME Type

Content-Type: application/llmfeed+json

Or specifically:

Content-Type: application/prompt+llmfeed

session.llmfeed.json: Captures prompt usage in workflows
credential.llmfeed.json: Authentication for premium prompts
capabilities.llmfeed.json: Services that use certified prompts
mcp.llmfeed.json: Overall service context including prompt libraries
workflow.llmfeed.json: Complete automation workflow definitions

References

Note: The combination of prompt.llmfeed.json with API orchestration platforms transforms static prompts into living, executable, and measurable AI workflows. This is where specification meets implementation, creating the foundation for the agentic economy.

Purpose​

The Evolution of Prompt Engineering​

Why JSON Goes Further​

Typical Use Cases​

Canonical Structure​

Core Fields​

🔌 API Execution & Orchestration​

From Specification to Execution​

API Integration Pattern​

Direct API Execution​

API Cost Tracking​

Batch Processing Support​

API Authentication Patterns​

Execution Monitoring​

🔄 Platform Integration Examples​

N8N Workflow Integration​

Zapier Integration​

Make (Integromat) Integration​

🚀 Real-World API Execution Pipeline​

Complete Execution Flow​

Healthcare-Specific API Considerations​

📊 Performance & Optimization​

API Response Caching​

Load Balancing Across APIs​

🔐 Security in API Execution​

API Key Management​

Trust & Ownership​

Basic Signing​

Certification (Optional)​

Agent Behavior​

Simple Examples​

Educational Prompt​

API Documentation Generator​

📚 Advanced Features​

Medical Consultation Assistant​

Legal Document Analysis​

Copyright Protection​

Commercial Licensing​

Performance Metrics​

LLM Compatibility​

Agent Guidance​

Trigger Targets​

Best Practices​

For Basic Users​

For Professional Use​

For Marketplaces​

For Automation Platforms​

MIME Type​

Related Feed Types​

References​