Contents
One of the most useful things in the BMad Method is the advanced elicitation system. When you’re deep in a document or design decision and you want to push the thinking further, BMad presents five context-matched techniques and lets you choose which angle to hit it from.
The full library has 50 methods. Most people only encounter five at a time through the shuffle interface. This post covers all of them — what each technique does, why it works, and where in the BMad workflow it earns its place.
The Commands
To start advanced elicitation directly:
/bmad-advanced-elicitation
You can also trigger it by mentioning a specific technique by name — say “run a pre-mortem on this”, “let’s do red team vs blue team”, or “apply first principles here” and BMad will invoke the elicitation skill automatically.
If you’re not sure where you are in a BMad session or what to do next:
/bmad-help
The help skill reads your current project state, figures out which phase you’re in, and tells you exactly what the next recommended step is — along with the command to run it. It’s the best starting point if you’re new to BMad or picking up a project mid-workflow.
How Advanced Elicitation Works in BMad
Any BMad agent can invoke the elicitation skill mid-session. When it runs, it analyzes the current content type and context, then selects five techniques it thinks fit. You choose one, the agent applies it to whatever you’re working on, and then the menu reappears. You can keep applying techniques in sequence, building on each enhancement.
The key insight: these aren’t separate prompts you copy-paste. They’re live analytical passes applied to your actual work in progress. Pre-mortem runs against your actual architecture. Red Team attacks your specific design. The Feynman Technique tries to explain your specific implementation simply enough to reveal the gaps.
Core Techniques
These six are the fundamentals — the ones that appear most often and apply in the broadest range of situations.
First Principles Analysis strips away inherited assumptions and rebuilds from fundamental truths. Most codebases carry decisions made for reasons nobody remembers. First principles is how you audit those decisions and separate the ones that are still load-bearing from the ones that are just convention. Use it in Phase 2 (Planning) when the architecture feels like cargo-culting.
5 Whys Deep Dive drills through surface explanations until it reaches root causes. Simple in theory, disarmingly hard in practice — most people stop at the first plausible answer. Use it in any phase when a problem keeps coming back, or when a solution isn’t holding.
Socratic Questioning probes assumptions with targeted questions rather than declarative critique. Instead of saying what’s wrong, it asks questions that lead you to discover what’s wrong yourself. Mary (Business Analyst) operates this way by default. Use it in Phase 1 when you suspect you’re not fully understanding your own requirements.
Critique and Refine is a structured review cycle — identify strengths, identify weaknesses, improve. The most straightforward of the core methods, but having it as an explicit step prevents the drift where you jump straight to refinement without properly cataloguing what’s actually working.
Explain Reasoning walks through the thinking step by step, showing how conclusions were reached. Crucial for transparency and for catching the logical jumps that felt natural in the moment but don’t actually hold. Use it when a decision needs to be documented or defended.
Expand or Contract for Audience adjusts the depth and technical density of content to match who’s reading it. Invaluable when what you’re writing will be read by people with different levels of technical background — engineers, clients, stakeholders, junior developers.
Collaboration Techniques
Ten methods that simulate multi-perspective input, each useful when a single viewpoint isn’t enough.
Stakeholder Round Table convenes multiple personas to contribute diverse perspectives. Essential for requirements gathering when you have competing interests. Output: perspectives → synthesis → alignment.
Expert Panel Review assembles domain experts for deep specialized analysis. Use when technical depth matters and you want something closer to peer review quality. Better than asking one generalist to cover a complex specialized domain.
Debate Club Showdown has two personas argue opposing positions while a moderator scores points. Great for controversial architectural decisions — it forces both sides to be articulated clearly before you pick one. The synthesis phase often produces better options than either starting position.
User Persona Focus Group gathers your product’s user personas to react to proposals and share frustrations. Essential for validating features before building them. Use in Phase 2 (Planning) when the product direction is still forming.
Time Traveler Council brings past-you and future-you into the conversation to advise present-you. Powerful for evaluating decisions that have long-term consequences that the present moment makes hard to see. Use when you’re about to make a commitment that’s hard to reverse.
Cross-Functional War Room runs a product manager, engineer, and designer through the same problem together. Reveals trade-offs between what’s feasible, what’s desirable, and what’s viable — the three constraints that live in different heads and rarely meet in one conversation.
Mentor and Apprentice has a senior expert explain something while a junior asks naive questions. The naive questions are the valuable part — they surface hidden assumptions through the act of teaching. When you can’t explain something simply to someone who doesn’t know it, that’s where your understanding has gaps.
Good Cop Bad Cop alternates a supportive persona and a critical persona. Finds both strengths to build on and weaknesses to address. Useful for anything that needs balanced assessment rather than pure advocacy or pure criticism.
Improv Yes-And builds ideas collaboratively without blocking — each persona adds to what came before. Generates unexpected creative directions. Use in early brainstorming when you need to break out of incremental thinking.
Customer Support Theater roleplays an angry customer and support rep to find pain points. Reveals real user frustrations and service gaps before they show up in production feedback.
Advanced Reasoning Techniques
Six methods that operate at the meta level — not just what to think, but how to think.
Tree of Thoughts explores multiple reasoning paths simultaneously, evaluates them, and selects the best. Perfect for complex problems with multiple valid approaches where the naive greedy path isn’t the right one.
Graph of Thoughts models reasoning as an interconnected network of ideas to reveal hidden relationships. Where Tree of Thoughts is sequential branching, Graph of Thoughts captures circular dependencies and emergent patterns — the kind of structure that shows up in distributed systems and domain models.
Thread of Thought maintains coherent reasoning across long contexts by weaving a continuous narrative thread. Essential for RAG systems and for any session where earlier context needs to stay active through later decisions.
Self-Consistency Validation generates multiple independent approaches then compares them for consistency. Use this for high-stakes decisions where you want verification — if three independent approaches reach the same conclusion, you have more confidence than if you just followed one chain of reasoning.
Meta-Prompting Analysis steps back to analyze the approach structure and methodology itself rather than the content. Use when the output feels wrong but you’re not sure why — often the prompt structure or the framing is the problem, not the underlying ideas.
Reasoning via Planning builds a reasoning tree guided by world models and goal states. Excellent for strategic planning and any problem where the steps need to be taken in a specific sequence with clear dependencies.
Competitive / Adversarial Techniques
Three methods that use structured conflict to find weaknesses before they find you.
Red Team vs Blue Team runs an adversarial attack-defend analysis to find vulnerabilities. The Red Team attacks; the Blue Team defends; the output is a hardened design with specific failure modes addressed. Critical for security-relevant decisions, and genuinely useful for any decision where you need to stress-test the reasoning. Use in Phase 3 (Solutioning) before committing to an architecture.
Shark Tank Pitch has an entrepreneur pitch to skeptical investors who poke holes. Stress-tests business viability and forces clarity on the value proposition. Use it when you’re building something that needs to justify its existence to stakeholders who aren’t already sold.
Code Review Gauntlet has senior developers with different philosophies review the same code. Surfaces style debates and finds consensus on best practices. Better than a single reviewer because it externalizes the tradeoffs that usually stay inside one person’s head.
Technical Techniques
Four methods designed specifically for software engineering problems.
Architecture Decision Records has multiple architect personas propose and debate architectural choices with explicit trade-offs. Ensures decisions are well-reasoned and documented — use in Phase 3 (Solutioning) for any decision you’ll need to explain or defend later.
Rubber Duck Debugging Evolved explains code to progressively more technical ducks until the bug appears. Forces clarity at multiple abstraction levels — the simple duck makes you explain what it’s supposed to do, the technical duck makes you explain why it’s failing. Combine with 5 Whys when the bug is systemic.
Algorithm Olympics has multiple approaches compete on the same problem with benchmarks. Use when there are genuinely multiple viable implementations and you need to pick one — benchmarks prevent the debate from being purely theoretical.
Security Audit Personas brings a hacker, defender, and auditor together to examine the system from different threat models. Comprehensive security review from multiple angles. The hacker finds what the auditor misses; the auditor finds what the defender rationalizes away. Use before any public launch.
Performance Profiler Panel has a database expert, frontend specialist, and DevOps engineer diagnose slowness together. Finds bottlenecks across the full stack — the ones that each specialist would miss because they’re in another layer.
Creative Techniques
Six methods for breaking out of linear thinking.
SCAMPER Method applies seven creativity lenses in sequence: Substitute, Combine, Adapt, Modify, Put to other uses, Eliminate, Reverse. Systematic ideation for product innovation. Use in Phase 2 when features feel incremental and you want to find genuinely different approaches.
Reverse Engineering works backwards from the desired outcome to find the implementation path. Powerful when the end state is clear but the path isn’t. Particularly useful for UX problems where you know what the user experience should feel like but not how to build it.
What If Scenarios explores alternative realities to understand possibilities and implications. Valuable for contingency planning — what if the third-party API goes down, what if usage scales 10x, what if the requirements change after launch.
Random Input Stimulus injects unrelated concepts to spark unexpected connections. Breaks creative blocks through forced lateral thinking. Use when you’ve been staring at the same problem for too long and every idea is a variation of the last.
Exquisite Corpse Brainstorm has each persona add to an idea seeing only the previous contribution. Generates surprising combinations through constrained collaboration. Useful for naming, marketing angles, and anything where too much context kills originality.
Genre Mashup combines two unrelated domains to find fresh approaches — what would this product be if it were designed like a video game, or a hotel, or a language school. Innovation through unexpected cross-pollination.
Research Techniques
Three methods for evaluating evidence and making structured comparisons.
Literature Review Personas pairs an optimist researcher, a skeptic researcher, and a synthesizer reviewing the same sources. Produces balanced assessment of evidence quality — the optimist finds what’s compelling, the skeptic finds what’s overreaching, the synthesizer finds the defensible middle.
Thesis Defense Simulation has a student defend a hypothesis against a committee with different concerns. Stress-tests research methodology and conclusions. Use in Phase 1 (Analysis) when you’re making assumptions about market or user behavior that need to be examined.
Comparative Analysis Matrix has multiple analysts evaluate options against weighted criteria. Structured decision-making with explicit scoring. Use when you have more than two meaningful options and need to make the comparison transparent rather than intuitive.
Risk Techniques
Five methods for anticipating and preventing failure.
Pre-mortem Analysis imagines a specific future failure and works backwards to understand what caused it. Reverses the psychological pressure that makes people underreport risks during planning — in a pre-mortem, pessimism is the assignment. Use before any major launch or architectural commit.
Failure Mode Analysis systematically explores how each component could fail. Critical for reliability engineering and safety-critical systems. More thorough than pre-mortem because it goes component by component rather than imagining a single failure scenario.
Challenge from Critical Perspective plays devil’s advocate to stress-test ideas and find weaknesses. Essential for overcoming groupthink — use when everyone in the session is already sold on the approach and you need someone to push back.
Identify Potential Risks brainstorms what could go wrong across all categories: technical, business, user, compliance, operational. Fundamental for project planning. More structured than pre-mortem — less narrative, more exhaustive.
Chaos Monkey Scenarios deliberately breaks things to test resilience and recovery. Ensures the system handles failures gracefully. Use before production deployment for any system that needs to be reliable under partial failure conditions.
Learning Techniques
Two methods focused on validating understanding rather than generating new ideas.
Feynman Technique explains a complex concept as simply as possible — as if teaching a child. The ultimate test of whether you actually understand something versus just recognizing the vocabulary. Where explanation breaks down, understanding has a gap. Use this on any design decision you’re about to commit to.
Active Recall Testing tests understanding without references to verify that knowledge is actually retained. Essential for identifying gaps — what you think you know versus what you can reconstruct from memory are often different things.
Philosophical Techniques
Two methods for navigating values and complexity.
Occam’s Razor Application finds the simplest sufficient explanation by eliminating unnecessary complexity. Essential for debugging and for architecture decisions where multiple explanations are available — the simplest one that fits the evidence is usually right until proven otherwise.
Trolley Problem Variations explores ethical trade-offs through moral dilemmas. Valuable for understanding values and making difficult decisions where multiple stakeholders have legitimate but competing interests. Use when you’re designing a system that makes choices with real consequences.
Retrospective Techniques
Two methods for learning from completed work.
Hindsight Reflection imagines looking back from the future to gain perspective on a completed project or decision. Powerful for project reviews — future-you can evaluate present decisions without the pressure that present-you is under.
Lessons Learned Extraction systematically identifies key takeaways and actionable improvements from an experience. Essential for continuous improvement — use at the end of every sprint, project, or major decision cycle. The output is specific actions, not vague intentions.
How to Use These in Practice
The BMad elicitation interface surfaces five context-matched techniques and lets you pick. When you type a in the elicitation menu, you get the full list of 50 with descriptions — you can then pick any by name.
A few patterns I’ve found useful:
For planning phases, lead with Pre-mortem, then First Principles if the plan needs to be rebuilt, then Stakeholder Round Table to find the competing interests.
For implementation review, Red Team vs Blue Team for security-relevant code, then Algorithm Olympics if there are multiple implementations worth comparing, then Rubber Duck Debugging Evolved if something isn’t working.
For stuck creative problems, start with Random Input Stimulus to break the frame, then What If Scenarios to explore the space, then SCAMPER to systematically cover what you might have missed.
For any decision you’re about to commit to, run the Feynman Technique first. If you can’t explain it simply, you don’t understand it well enough to build it yet.
The full BMad Method methodology and installation instructions are at docs.bmad-method.org. The overview of the framework and its agents is in my introduction to the BMad Method.
If you run party mode sessions, Agent Vibes gives each BMad agent a distinct voice — useful when multiple agents are applying elicitation techniques simultaneously and you want to track who’s speaking without reading response headers. For the full picture on Claude Code and the tools these techniques apply to, the AI tools roundup for 2026 covers benchmarks, plan tiers, and the broader ecosystem.
Comments
Loading comments…
Leave a comment