pig-farm-controller/bmad/bmm/workflows/1-analysis/research/instructions-deep-prompt.md

Deep Research Prompt Generator Instructions

The workflow execution engine is governed by: {project_root}/bmad/core/tasks/workflow.xml You MUST have already loaded and processed: {installed_path}/workflow.yaml This workflow generates structured research prompts optimized for AI platforms Based on 2025 best practices from ChatGPT, Gemini, Grok, and Claude

Understand what the user wants to research

Let's create a powerful deep research prompt!

What topic or question do you want to research?

Examples:

• "Future of electric vehicle battery technology"
• "Impact of remote work on commercial real estate"
• "Competitive landscape for AI coding assistants"
• "Best practices for microservices architecture in fintech"

research_topic

What's your goal with this research?

• Strategic decision-making
• Investment analysis
• Academic paper/thesis
• Product development
• Market entry planning
• Technical architecture decision
• Competitive intelligence
• Thought leadership content
• Other (specify)

research_goal

Which AI platform will you use for the research?

1. ChatGPT Deep Research (o3/o1)
2. Gemini Deep Research
3. Grok DeepSearch
4. Claude Projects
5. Multiple platforms
6. Not sure yet

target_platform

Help user define clear boundaries for focused research

Let's define the scope to ensure focused, actionable results:

Temporal Scope - What time period should the research cover?

• Current state only (last 6-12 months)
• Recent trends (last 2-3 years)
• Historical context (5-10 years)
• Future outlook (projections 3-5 years)
• Custom date range (specify)

temporal_scope

Geographic Scope - What geographic focus?

• Global
• Regional (North America, Europe, Asia-Pacific, etc.)
• Specific countries
• US-focused
• Other (specify)

geographic_scope

Thematic Boundaries - Are there specific aspects to focus on or exclude?

Examples:

• Focus: technological innovation, regulatory changes, market dynamics
• Exclude: historical background, unrelated adjacent markets

thematic_boundaries

Determine what types of information and sources are needed

What types of information do you need?

Select all that apply:

• Quantitative data and statistics
• Qualitative insights and expert opinions
• Trends and patterns
• Case studies and examples
• Comparative analysis
• Technical specifications
• Regulatory and compliance information
• Financial data
• Academic research
• Industry reports
• News and current events

information_types

Preferred Sources - Any specific source types or credibility requirements?

Examples:

• Peer-reviewed academic journals
• Industry analyst reports (Gartner, Forrester, IDC)
• Government/regulatory sources
• Financial reports and SEC filings
• Technical documentation
• News from major publications
• Expert blogs and thought leadership
• Social media and forums (with caveats)

preferred_sources

Specify desired output format for the research

Output Format - How should the research be structured?

1. Executive Summary + Detailed Sections
2. Comparative Analysis Table
3. Chronological Timeline
4. SWOT Analysis Framework
5. Problem-Solution-Impact Format
6. Question-Answer Format
7. Custom structure (describe)

output_format

Key Sections - What specific sections or questions should the research address?

Examples for market research:

• Market size and growth
• Key players and competitive landscape
• Trends and drivers
• Challenges and barriers
• Future outlook

Examples for technical research:

• Current state of technology
• Alternative approaches and trade-offs
• Best practices and patterns
• Implementation considerations
• Tool/framework comparison

key_sections

Depth Level - How detailed should each section be?

• High-level overview (2-3 paragraphs per section)
• Standard depth (1-2 pages per section)
• Comprehensive (3-5 pages per section with examples)
• Exhaustive (deep dive with all available data)

depth_level

Gather additional context to make the prompt more effective

Persona/Perspective - Should the research take a specific viewpoint?

Examples:

• "Act as a venture capital analyst evaluating investment opportunities"
• "Act as a CTO evaluating technology choices for a fintech startup"
• "Act as an academic researcher reviewing literature"
• "Act as a product manager assessing market opportunities"
• No specific persona needed

research_persona

Special Requirements or Constraints:

• Citation requirements (e.g., "Include source URLs for all claims")
• Bias considerations (e.g., "Consider perspectives from both proponents and critics")
• Recency requirements (e.g., "Prioritize sources from 2024-2025")
• Specific keywords or technical terms to focus on
• Any topics or angles to avoid

special_requirements

{project-root}/bmad/core/tasks/adv-elicit.xml

Establish how to validate findings and what follow-ups might be needed

Validation Criteria - How should the research be validated?

• Cross-reference multiple sources for key claims
• Identify conflicting viewpoints and resolve them
• Distinguish between facts, expert opinions, and speculation
• Note confidence levels for different findings
• Highlight gaps or areas needing more research

validation_criteria

Follow-up Questions - What potential follow-up questions should be anticipated?

Examples:

• "If cost data is unclear, drill deeper into pricing models"
• "If regulatory landscape is complex, create separate analysis"
• "If multiple technical approaches exist, create comparison matrix"

follow_up_strategy

Synthesize all inputs into platform-optimized research prompt

Generate the deep research prompt using best practices for the target platform

Prompt Structure Best Practices:

1. Clear Title/Question (specific, focused)
2. Context and Goal (why this research matters)
3. Scope Definition (boundaries and constraints)
4. Information Requirements (what types of data/insights)
5. Output Structure (format and sections)
6. Source Guidance (preferred sources and credibility)
7. Validation Requirements (how to verify findings)
8. Keywords (precise technical terms, brand names)

Generate prompt following this structure

deep_research_prompt

Review the generated prompt:

• [a] Accept and save
• [e] Edit sections
• [r] Refine with additional context
• [o] Optimize for different platform

What would you like to adjust? Regenerate with modifications Provide platform-specific usage tips based on target platform **ChatGPT Deep Research Tips:**

• Use clear verbs: "compare," "analyze," "synthesize," "recommend"
• Specify keywords explicitly to guide search
• Answer clarifying questions thoroughly (requests are more expensive)
• You have 25-250 queries/month depending on tier
• Review the research plan before it starts searching

**Gemini Deep Research Tips:**

• Keep initial prompt simple - you can adjust the research plan
• Be specific and clear - vagueness is the enemy
• Review and modify the multi-point research plan before it runs
• Use follow-up questions to drill deeper or add sections
• Available in 45+ languages globally

**Grok DeepSearch Tips:**

• Include date windows: "from Jan-Jun 2025"
• Specify output format: "bullet list + citations"
• Pair with Think Mode for reasoning
• Use follow-up commands: "Expand on [topic]" to deepen sections
• Verify facts when obscure sources cited
• Free tier: 5 queries/24hrs, Premium: 30/2hrs

**Claude Projects Tips:**

• Use Chain of Thought prompting for complex reasoning
• Break into sub-prompts for multi-step research (prompt chaining)
• Add relevant documents to Project for context
• Provide explicit instructions and examples
• Test iteratively and refine prompts

platform_tips

Create a checklist for executing and evaluating the research

Generate execution checklist with:

Before Running Research:

• Prompt clearly states the research question
• Scope and boundaries are well-defined
• Output format and structure specified
• Keywords and technical terms included
• Source guidance provided
• Validation criteria clear

During Research:

• Review research plan before execution (if platform provides)
• Answer any clarifying questions thoroughly
• Monitor progress if platform shows reasoning process
• Take notes on unexpected findings or gaps

After Research Completion:

• Verify key facts from multiple sources
• Check citation credibility
• Identify conflicting information and resolve
• Note confidence levels for findings
• Identify gaps requiring follow-up
• Ask clarifying follow-up questions
• Export/save research before query limit resets

execution_checklist

Save complete research prompt package

Your Deep Research Prompt Package is ready!

The output includes:

1. Optimized Research Prompt - Ready to paste into AI platform
2. Platform-Specific Tips - How to get the best results
3. Execution Checklist - Ensure thorough research process
4. Follow-up Strategy - Questions to deepen findings

Save all outputs to {default_output_file}

Would you like to:

1. Generate a variation for a different platform
2. Create a follow-up prompt based on hypothetical findings
3. Generate a related research prompt
4. Exit workflow

Select option (1-4):

Start with different platform selection Start new prompt with context from previous Load the FULL file: {output_folder}/bmm-workflow-status.yaml Find workflow_status key "research" ONLY write the file path as the status value - no other text, notes, or metadata Update workflow_status["research"] = "{output_folder}/bmm-research-deep-prompt-{{date}}.md" Save file, preserving ALL comments and structure including STATUS DEFINITIONS

Find first non-completed workflow in workflow_status (next workflow to do) Determine next agent from path file based on next workflow

✅ Deep Research Prompt Generated

Research Prompt:

• Structured research prompt generated and saved to {output_folder}/bmm-research-deep-prompt-{{date}}.md
• Ready to execute with ChatGPT, Claude, Gemini, or Grok

{{#if standalone_mode != true}} Status Updated:

• Progress tracking updated: research marked complete
• Next workflow: {{next_workflow}} {{else}} Note: Running in standalone mode (no progress tracking) {{/if}}

Next Steps:

{{#if standalone_mode != true}}

• Next workflow: {{next_workflow}} ({{next_agent}} agent)
• Optional: Execute the research prompt with AI platform, gather findings, or run additional research workflows

Check status anytime with: workflow-status {{else}} Since no workflow is in progress:

• Execute the research prompt with AI platform and gather findings
• Refer to the BMM workflow guide if unsure what to do next
• Or run workflow-init to create a workflow path and get guided next steps {{/if}}