Detailed reference for all seven Vortex agents, their workflows, positioning, and how they work together in the complete 7-stream product discovery pattern. The Innovation Vortex is a pattern from the unFIX model by Jurgen Appelo.
Stream: Contextualize | Icon: 🎯 | File: _bmad/bme/_vortex/agents/contextualization-expert.md
Emma helps teams frame the right problem before building solutions. She focuses on Lean Startup methodologies, validated assumptions, and strategic clarity.
| Cmd | Workflow | Steps | What it produces |
|---|---|---|---|
| LP | Lean Persona | 6 | Jobs-to-be-done focused persona with riskiest assumptions identified |
| PV | Product Vision | 6 | Vision statement, future state (3-5 years), guiding principles, validation plan |
| CS | Contextualize Scope | 6 | Scope decision with scoring matrix, clear boundaries, and rationale |
- You're starting a new product or feature and need to define the problem space
- You want to create hypothesis-driven personas grounded in research
- You need strategic alignment before tactical execution
Emma vs Maya (BMAD Core):
- Emma contextualizes — "What problem should we solve?"
- Maya creates — "How should we solve it?"
Emma answers the strategic question before Maya answers the tactical one.
Stream: Empathize | Icon: 🔍 | File: _bmad/bme/_vortex/agents/discovery-empathy-expert.md
Isla helps teams deeply understand users before designing solutions. She focuses on evidence-based user insights over assumptions, bridging strategic framing and experimentation.
| Cmd | Workflow | Steps | What it produces |
|---|---|---|---|
| EM | Empathy Map | 6 | Empathy map artifact with says/thinks/does/feels quadrants and pain/gain analysis |
| UI | User Interview | 6 | Interview script with research goals, findings capture, and pattern synthesis |
| UD | User Discovery | 6 | Discovery research report with research planning, execution, and key findings |
- You know the problem area but need to understand your users better
- You want evidence-based empathy maps instead of assumption-based ones
- You're planning user interviews and need structured research goals
Isla vs Emma:
- Isla empathizes — "Who are our users and what do they need?"
- Emma contextualizes — "What problem space are we in?"
Isla provides the user understanding that informs Emma's strategic framing.
Stream: Synthesize | Icon: 🔬 | File: _bmad/bme/_vortex/agents/research-convergence-specialist.md
Mila converges divergent research streams into actionable problem definitions. She transforms raw empathy data and contextual insights into clear, prioritized problem statements using Jobs-to-be-Done framing and Pains & Gains analysis.
| Cmd | Workflow | Steps | What it produces |
|---|---|---|---|
| RC | Research Convergence | 5 | Single problem definition synthesized from divergent research artifacts |
| PR | Pivot Resynthesis | 5 | Re-synthesized problem definition incorporating evidence from failed experiments |
| PA | Pattern Mapping | 5 | Cross-source pattern map surfacing convergent themes across research |
- You have multiple research artifacts from Isla and need a single problem definition
- An experiment failed and you need to re-synthesize the problem with new evidence
- You have data from multiple sources and want to surface convergent patterns
Mila vs Isla:
- Isla empathizes — "Let me understand the users deeply"
- Mila synthesizes — "Here's what all that research means together"
Mila turns Isla's divergent research into a converged, actionable problem definition.
Stream: Hypothesize | Icon: 💡 | File: _bmad/bme/_vortex/agents/hypothesis-engineer.md
Liam engineers testable hypotheses from validated problem definitions. He ideates alongside the user as a creative peer, using structured brainwriting, 4-field hypothesis contracts, and assumption mapping to turn problems into experiments.
| Cmd | Workflow | Steps | What it produces |
|---|---|---|---|
| HE | Hypothesis Engineering | 5 | Testable hypotheses in 4-field contract format with riskiest assumptions identified |
| AM | Assumption Mapping | 4 | Classified assumptions by risk level (lethality x uncertainty) |
| ED | Experiment Design | 4 | Experiment designs targeting riskiest assumptions first |
- You have a validated problem definition and need testable solution hypotheses
- You want to surface and classify hidden assumptions before building
- You need to design experiments that target the riskiest assumptions, not the easiest
Liam vs Wade:
- Liam hypothesizes — "What should we test and why?"
- Wade externalizes — "Let's run the test and measure"
Liam produces the hypothesis contracts that Wade executes as experiments.
Stream: Externalize | Icon: 🧪 | File: _bmad/bme/_vortex/agents/lean-experiments-specialist.md
Wade helps teams test assumptions with real users through rapid experiments. He focuses on validated learning over perfection, guiding teams from hypotheses to evidence.
| Cmd | Workflow | Steps | What it produces |
|---|---|---|---|
| MVP | Minimum Viable Product | 6 | MVP specification — the smallest thing to test an assumption (not a feature-light product) |
| LE | Lean Experiment | 6 | Build-Measure-Learn cycle with hypothesis, metrics, and pivot-or-persevere criteria |
| POC | Proof-of-Concept | 6 | Technical feasibility evaluation — "Can we build it?" |
| POV | Proof-of-Value | 6 | Business value validation — willingness to pay, market demand, build/pivot/kill decision |
- You have hypotheses ready to test with real users
- You need to validate technical feasibility before committing resources
- You want to test business value before building
Wade vs Sally (BMAD Core):
- Wade externalizes — "Should we build this?" (test with users)
- Sally internalizes — "Are we building it well?" (test with code)
Wade validates the product direction before Sally validates the implementation.
Stream: Sensitize | Icon: 📡 | File: _bmad/bme/_vortex/agents/production-intelligence-specialist.md
Noah interprets production signals through the lens of experiment history. He connects real-world metrics to their original hypotheses, revealing what production usage actually means for product-market fit. Noah reports — he does not prescribe strategy.
| Cmd | Workflow | Steps | What it produces |
|---|---|---|---|
| SI | Signal Interpretation | 5 | Contextual interpretation of a production signal within its experiment lineage |
| BA | Behavior Analysis | 5 | Behavioral pattern analysis against validated experiment baselines |
| MO | Production Monitoring | 5 | Multi-signal monitoring dashboard across active experiments |
- An experiment graduated to production and you need to interpret the signals
- You want to detect behavioral patterns that surveys and dashboards miss
- You're monitoring multiple experiments in production and need holistic intelligence
Noah vs Max:
- Noah sensitizes — "Here's what the data shows in context"
- Max systematizes — "Here's what we should do about it"
Noah provides the intelligence; Max makes the strategic decisions.
Stream: Systematize | Icon: 🧭 | File: _bmad/bme/_vortex/agents/learning-decision-expert.md
Max helps teams systematize learnings and make evidence-based decisions. He captures validated learning as organizational knowledge and navigates between Vortex streams based on evidence gaps.
| Cmd | Workflow | Steps | What it produces |
|---|---|---|---|
| LC | Learning Card | 6 | Structured learning capture with experiment context, results, and implications |
| PPP | Pivot, Patch, or Persevere | 6 | Evidence-based decision with quantitative thresholds and rationale |
| VN | Vortex Navigation | 6 | Stream recommendation based on gap analysis across all Vortex streams |
- You've completed experiments and need to document what was learned
- You need to make a pivot/persevere decision based on evidence
- You're unsure which Vortex stream to focus on next
Max vs Wade:
- Max systematizes — "What did we learn? What's next?"
- Wade externalizes — "Let's test this assumption"
Max turns Wade's experiment results into organizational knowledge and strategic direction.
The Vortex is non-linear by design. While there's a natural forward flow (Isla → Mila → Liam → Wade → Noah → Max), agents route back to each other based on evidence. Ten handoff contracts ensure structured information flows at every transition.
VORTEX PATTERN — 7 Streams · 7 Agents
┌──────────┐ HC1 ┌──────────┐ HC2 ┌──────────┐ HC3 ┌──────────┐
│ Isla 🔍 │─────▶│ Mila 🔬 │─────▶│ Liam 💡 │─────▶│ Wade 🧪 │
│ Empathize │ │Synthesize│ │Hypothesiz│ │Externaliz│
└──────────┘ └──────────┘ └──────────┘ └──────────┘
▲ ▲ │ │
│ HC6 │ HC9│ HC4│
│ │ ▼ ▼
┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐
│ Emma 🎯 │◀HC8─│ Max 🧭 │◀HC5─│ Noah 📡 │◀────────┘
│Contextual│ │Systematiz│ │ Sensitize│
└──────────┘ └──────────┘ └──────────┘
│ HC7 │ HC10│
└──────────────────┴──────────────────┘
▼ to Isla 🔍
You can start with any agent. Max's Vortex Navigation workflow is the compass that helps you decide where to go based on what evidence you have and what gaps remain.
The Vortex uses 10 handoff contracts (HC1-HC10) to ensure agents pass structured information to each other. There are three contract types:
These produce schema-compliant files that the next agent requires as input. Contract schemas are in _bmad/bme/_vortex/contracts/.
| Contract | Flow | What gets passed |
|---|---|---|
| HC1 | Isla 🔍 → Mila 🔬 | Empathy artifacts (maps, interviews, observations) with synthesized insights |
| HC2 | Mila 🔬 → Liam 💡 | Converged problem definition (JTBD + Pains & Gains) with evidence summary |
| HC3 | Liam 💡 → Wade 🧪 | 1-3 hypothesis contracts in 4-field format with risk map |
| HC4 | Wade 🧪 → Noah 📡 | Graduated experiment context (results, success criteria, metrics) |
| HC5 | Noah 📡 → Max 🧭 | Signal report (signal + experiment lineage + trend analysis) |
Max makes strategic routing decisions based on evidence. No artifact file — context is carried in conversation.
| Contract | Flow | When it triggers |
|---|---|---|
| HC6 | Max 🧭 → Mila 🔬 | Pivot decision: problem correct, solution direction failed. Mila re-synthesizes. |
| HC7 | Max 🧭 → Isla 🔍 | Evidence gap: critical knowledge gap needs discovery research |
| HC8 | Max 🧭 → Emma 🎯 | Recontextualization: the fundamental problem space is wrong or scope too narrow |
Mid-workflow flags that appear in the Compass when triggered.
| Contract | Flow | When it triggers |
|---|---|---|
| HC9 | Liam 💡 → Isla 🔍 | Liam flags an unvalidated assumption too risky to test without prior validation |
| HC10 | Noah 📡 → Isla 🔍 | Noah detects unexpected user behavior not covered by the original hypothesis |
Every workflow's final step includes a Vortex Compass — a routing table that recommends what to do next based on what you just learned:
| If you learned... | Consider next... | Agent | Why |
|----------------------------|----------------------|----------|----------------------------|
| [evidence/condition] | [workflow-name] | Agent 🔍 | [rationale + HC reference] |
The Compass is guidance, not a mandate. You can navigate to any Vortex agent at any time based on your judgment. When you're unsure, run Max's Vortex Navigation for a full gap analysis across all 7 streams.
These journeys show how practitioners move through the Vortex in practice.
You have a pile of research artifacts from Isla but no single problem definition.
- Isla 🔍 conducts discovery — empathy maps, interviews, observations → produces multiple research artifacts
- Mila 🔬 (HC1) receives all artifacts → guides JTBD framing and Pains & Gains analysis across sources → produces a single converged problem definition
- Compass routes to Liam 💡 (HC2) for hypothesis engineering, or back to Isla 🔍 if assumptions need validation first
You have a validated problem definition but no idea what solution to test first.
- Mila 🔬 produces a converged problem definition (e.g., "users waste 40 minutes daily context-switching")
- Liam 💡 (HC2) receives the problem → runs structured brainwriting → produces 3 hypothesis contracts with 4-field format (belief, evidence needed, experiment, success criteria) → identifies riskiest assumptions
- Compass routes to Wade 🧪 (HC3) with hypothesis contracts. If Liam flagged an unvalidated assumption → also offers Isla 🔍 (HC9) for quick validation
Your experiment graduated to production but the metrics are diverging from experiment results.
- Wade 🧪 validates an experiment (72% onboarding completion vs 60% threshold) → experiment graduates to production
- Noah 📡 (HC4) inherits experiment context → interprets production signals ("54% completion, down from validated 72%") → produces signal report in signal + context + trend format
- Max 🧭 (HC5) receives intelligence → captures learning or triggers pivot/patch/persevere decision
- If pivot → Mila 🔬 (HC6) re-synthesizes. If evidence gap → Isla 🔍 (HC7) investigates.
You're new to the Vortex with production data but no experiment history.
- Max 🧭 runs Vortex Navigation → analyzes evidence across all 7 streams → identifies widest gaps
- Max determines: "You have production signals but no experiment context (Sensitize gap), no hypothesis contracts (Hypothesize gap), and no converged problem definition (Synthesize gap)"
- Compass routes to the agent that fills the most critical gap — there is no mandatory "step 1." The Vortex meets you where your evidence is.
Max says "pivot" — the solution is wrong but the problem is right.
- Max 🧭 delivers pivot decision via Pivot/Patch/Persevere analysis
- Mila 🔬 (HC6) receives Isla's original artifacts + new evidence from failed experiments → re-synthesizes pains and gains while preserving the JTBD
- Liam 💡 (HC2) receives updated problem definition → engineers new hypotheses grounded in revised understanding
- The Vortex loops within the known problem space — it doesn't restart from scratch
You want to add a custom agent to the Vortex.
Use BMB (BMAD Module Builder) — the recommended way to extend BMAD-Enhanced. Run /bmad-bmb-agent to create a new agent through a guided process, or /bmad-bmb-module for a complete module with multiple agents and workflows. See the extension guidance FAQ for details.
Under the hood, BMB generates the 4 components every agent needs:
- Registry entry — In
scripts/update/lib/agent-registry.js(single source of truth for all agents and workflows) - Agent file — Agent definition in
_bmad/bme/_vortex/agents/(persona, communication style, menu) - Workflows — Workflow directories in
_bmad/bme/_vortex/workflows/with step files (4-6 steps each, final step includes Compass routing) - User guide — Guide in
_bmad/bme/_vortex/guides/with examples and artifact templates
The installer, validator, and doctor all read from the registry automatically — no additional wiring needed.
You're a team lead reviewing your team's Vortex work.
All generated artifacts are saved to _bmad-output/vortex-artifacts/. Each artifact is self-documenting:
- Problem definitions (Mila) reference their input Isla artifacts
- Hypothesis contracts (Liam) reference their source problem definition
- Experiment results (Wade) reference the hypothesis contract they tested
- Signal reports (Noah) reference the experiment that produced the production metrics
- Decision records (Max) reference the signals and learnings that drove the decision
You can trace any decision back to its evidence chain without using the agents directly.
All generated artifacts are saved to _bmad-output/vortex-artifacts/.
Each agent also has a user guide:
- Emma User Guide
- Isla User Guide
- Mila User Guide
- Liam User Guide
- Wade User Guide
- Noah User Guide
- Max User Guide
Each agent works standalone — you don't need all seven:
- Just Emma for strategic framing and problem-solution fit
- Just Isla for user research and empathy mapping
- Just Mila for converging research into problem definitions
- Just Liam for hypothesis engineering and assumption mapping
- Just Wade for experiment design and validation
- Just Noah for production signal interpretation
- Just Max for learning capture and decision-making
Use all seven together for the complete Vortex flow, or any subset that fits your needs.
Back to README | Testing | Development | FAQ