How to Use PMCR-O Prompts

The PMCR-O Prompt Library contains production-ready agent prompts designed for portability across AI runtimes. This guide shows you how to copy, customize, and deploy them effectively.

Step 1: Navigate to the Prompt Library

Visit pmcro-prompt-library.html and browse the agent cards. Each card contains:

• Agent Description: Role, purpose, and usage context
• Full Prompt: Click "Show Prompt" to reveal the complete template
• Copy Button: One-click copy of the entire prompt

Step 2: Choose Your Agent

Select an agent based on your task:

• Meta-Orchestrator: For complex, multi-phase projects
• Blog Writer Agent: For SEO-optimized articles
• Prompt Factory Agent: For creating new agent templates
• Reflection Mechanism Agent: For intent refinement
• Foundation Specialist: For architecture decisions

Step 3: Copy the Prompt

Click the "Copy" button on the agent card. The entire prompt template is copied to your clipboard, including:

• @meta block (agent metadata)
• @identity block (first-person declarations)
• @capabilities block (tool declarations)
• @context_template block (customizable placeholders)
• @constraints block (MANDATORY/FORBIDDEN rules)
• @bip_logic block (self-verification checks)
• @interaction block (usage instructions)

Step 4: Customize the Context Template

Every prompt includes a @context_template with placeholders. Replace them with your specific values:

@context_template {
  "seed_intent": "[INSERT_SEED_INTENT]",  ← Replace with your actual intent
  "target_runtime": "[Cursor | ChatGPT | Grok | Claude | Other]",  ← Your AI tool
  "tooling_available": ["[web_search? browser? none?]"],  ← What tools you have
  "success_criteria": ["[INSERT_MEASURABLE_OUTCOMES]"]  ← How to measure success
}

Example: Customizing the Blog Writer Agent

@context_template {
  "topic": "Behavioral Intent Programming",
  "target_audience": "Senior .NET developers",
  "target_runtime": "Cursor",
  "tooling_available": ["web_search", "browser_navigate"],
  "primary_conversion": "GitHub stars and forks",
  "internal_links": ["/bip-manifesto.html", "/pmcro-codex.html"]
}

Step 5: Set Target Runtime

PMCR-O prompts are portable, but you must specify your runtime:

• Cursor: Has file system access, code editing, web search
• ChatGPT: Has web search (if enabled), no file system
• Grok: Has web search, X integration
• Claude: Has web search (if enabled), no file system

Step 6: Paste and Execute

Paste the customized prompt into your AI tool and provide your seed intent:

[Paste full prompt here]

USER INPUT: Build a PMCR-O agent service with PostgreSQL knowledge vault for RAG.

Step 7: Understand the BIP Logic

Every prompt includes @bip_logic—self-verification checks the agent performs:

• CHECK 1: Confirms target_runtime and tooling_available
• CHECK 2: Restates your intent in one sentence
• CHECK 3: Chooses the next PMCR-O phase (Planner/Maker/Checker/Reflector)
• CHECK 4: Cites sources if tools were used, or states "No external validation performed"
• CHECK 5: Produces artifacts (files, code, checklists) not just advice

Common Use Cases

Use Case 1: Writing an Article

1. Copy the Blog Writer Agent v1.1 prompt
2. Set topic, target_audience, target_runtime
3. Paste into Cursor or ChatGPT
4. Provide seed intent: "Write an article about [TOPIC]"
5. The agent outputs: Title tag, meta description, canonical URL, OpenGraph tags, Twitter cards, Schema.org JSON-LD, full article markdown, internal link plan

Use Case 2: Creating a New Agent

1. Copy the Prompt Factory Agent v1.1 prompt
2. Set categories (e.g., ["Code Review", "Documentation"])
3. Set target_runtime and tooling_available
4. Paste and provide: "GENERATE ASSETS for [CATEGORIES]"
5. The agent outputs: Library Index + 3-7 reusable templates per category

Use Case 3: Refining Intent

1. Copy the Reflection Mechanism Agent v1.1 prompt
2. Set target_runtime and tooling_available
3. Paste and provide your raw intent
4. The agent mirrors it back as refined truth with validation

Best Practices

• Always set target_runtime: This enables proper tool-gating
• Be specific with tooling_available: List actual tools (web_search, browser_navigate) or ["none"]
• Provide measurable success_criteria: "Article ranks on page 1 for 'BIP prompt engineering'"
• Use the identity-first structure: Don't remove the "I AM" declarations—they create agency
• Respect the constraints: The MANDATORY/FORBIDDEN rules prevent common failures

Troubleshooting

Agent Claims It Searched But Didn't

This violates the "Evidence or Disclaimer" rule. Check that:

• target_runtime is set correctly
• tooling_available accurately reflects your tools
• The prompt includes the tool-gating rule: "If tools unavailable, say No external validation performed."

Agent Produces Advice Instead of Artifacts

The @bip_logic CHECK 5 requires artifacts. If you get advice, remind the agent: "Produce artifacts (files, code, checklists) not advice."

Agent Skips Verification Checks

The @bip_logic block is mandatory. If checks are skipped, paste the prompt again and explicitly request: "Execute all @bip_logic checks."

Next Steps

• Read BIP Prompt Engineering Guide for advanced customization
• Explore the complete prompt library
• Learn how to build your own agents