11-agent evidence-based research pipeline: from question formulation to bias-audited synthesis, every conclusion earns its confidence level.
Version: 1.0.0 | Created: 2026-02-07 | Total: 11 agents, 5,386 lines
dr-orchestrator (Pipeline Coordinator)
[Orchestrator]
|
┌────────────────────┼────────────────────┐
| | |
┌────────────────┐ ┌──────────────┐ ┌──────────────┐
| TIER 0 | | TIER 1 | | QA |
| Diagnostic | | Execution | | Audit |
| (sequential) | | (parallel) | | (sequential)|
└────────────────┘ └──────────────┘ └──────────────┘
| | |
┌───────────┐ ┌───────────────┐ ┌───────────┐
| Sackett | | Forsgren | | Ioannidis |
| PICO | | DORA/SPACE | | Evidence |
└─────┬─────┘ ├───────────────┤ | Reliability|
| | Cochrane | └─────┬─────┘
┌─────┴─────┐ | PRISMA | |
| Booth | ├───────────────┤ ┌─────┴─────┐
| SALSA | | Higgins | | Kahneman |
└─────┬─────┘ | OSINT | | Cognitive |
| ├───────────────┤ | Bias Audit|
┌─────┴─────┐ | Klein | └───────────┘
| Creswell | | NDM/RPD |
| Mixed | ├───────────────┤
| Methods | | Gilad |
└───────────┘ | CI/SCIP |
└───────────────┘
Flow: Query → Classification → Tier 0 → Tier 1 → QA → Synthesis → Report
| Tier | Agent | Persona | Based On | Lines | Focus |
|---|---|---|---|---|---|
| Orch | dr-orchestrator | DR Orchestrator | Original | 541 | Query classification, pipeline routing, quality gates |
| 0 | sackett | David Sackett | EBM founder | 595 | PICO question formulation, evidence levels |
| 0 | booth | Andrew Booth | SALSA framework | 682 | Review methodology selection (14 types) |
| 0 | creswell | John Creswell | Mixed methods pioneer | 698 | Qual/quant/mixed research design |
| 1 | forsgren | Nicole Forsgren | DORA/SPACE | 363 | Technical performance measurement |
| 1 | cochrane | Cochrane Collaboration | Systematic reviews | 456 | Evidence synthesis, PRISMA, GRADE |
| 1 | higgins | Eliot Higgins | Bellingcat/OSINT | 367 | Open-source investigations, verification |
| 1 | klein | Gary Klein | NDM/RPD | 430 | Pattern recognition, sensemaking |
| 1 | gilad | Ben Gilad | Competitive Intelligence | 385 | CI, strategic early warning, blind spots |
| QA | ioannidis | John Ioannidis | Meta-research | 408 | Evidence reliability, PPV, bias detection |
| QA | kahneman | Daniel Kahneman | Behavioral Economics | 461 | Cognitive bias audit, decision hygiene |
| ID | Use Case | Primary Agents | Secondary Agents | When to Use |
|---|---|---|---|---|
| UC-001 | Technical Deep Dive | Forsgren + Cochrane | Klein, Higgins | Performance metrics, tech stack evaluation, DevOps assessment |
| UC-002 | Strategic Decision Support | Klein + Gilad | Forsgren, Cochrane | Build vs buy, architecture decisions, strategic pivots |
| UC-003 | Competitive Intelligence | Gilad + Higgins | Klein | Market analysis, competitor tracking, early warning signals |
| UC-004 | Evidence Synthesis | Cochrane + Forsgren | Higgins | Literature reviews, state-of-the-art surveys, best practices |
@deep-research:dr-orchestrator
Or use the AIOX activation:
/AIOX:agents:dr-orchestrator
/AIOX:agents:sackett # PICO question formulation
/AIOX:agents:booth # Review methodology selection
/AIOX:agents:creswell # Research design (qual/quant/mixed)
/AIOX:agents:forsgren # Technical metrics (DORA/SPACE)
/AIOX:agents:cochrane # Systematic reviews (PRISMA)
/AIOX:agents:higgins # OSINT investigations
/AIOX:agents:klein # Pattern recognition / sensemaking
/AIOX:agents:gilad # Competitive intelligence
/AIOX:agents:ioannidis # Evidence reliability audit
/AIOX:agents:kahneman # Cognitive bias audit
# UC-001: Technical Deep Dive
"What are the performance implications of migrating from REST to gRPC for our microservices?"
# UC-002: Strategic Decision Support
"Should we build our own auth system or adopt Auth0/Clerk?"
# UC-003: Competitive Intelligence
"Map the AI code assistant landscape: Cursor, Windsurf, Claude Code, GitHub Copilot."
# UC-004: Evidence Synthesis
"What does the evidence say about monorepo vs polyrepo for teams of 20-50 engineers?"
| ID | Gate | Agent | Placement | Criteria | On Failure |
|---|---|---|---|---|---|
| QG-001 | Question Quality | Sackett | After Tier 0 | PICO question well-formed with all 4 components | Return to Sackett for reformulation |
| QG-002 | Methodology Fit | Booth | After Tier 0 | Review type matches question type and available evidence | Booth re-selects methodology |
| QG-003 | Evidence Reliability | Ioannidis | After Tier 1 | PPV > threshold, bias patterns flagged, unreliable evidence identified | Flag low-confidence findings |
| QG-004 | Decision Quality | Kahneman | Final | 12-Question Checklist passed, cognitive biases audited | Pre-mortem required before approval |
This squad was created through iterative research with devil's advocate validation. Each agent is based on real researchers, frameworks, or organizations with documented methodologies:
| Mind | Contribution | Domain |
|---|---|---|
| David Sackett | Evidence-Based Medicine founder, PICO framework, 5-step EBM | Clinical epidemiology |
| Andrew Booth | SALSA framework, STARLITE, review type classification (14 types) | Information science |
| John Creswell | Mixed methods research design, qual/quant integration strategies | Research methodology |
| Nicole Forsgren | DORA metrics, SPACE framework, Accelerate book | DevOps research |
| Cochrane Collaboration | Systematic review standards, PRISMA, GRADE evidence hierarchy | Healthcare research |
| Eliot Higgins | Bellingcat, open-source investigation methodology, OSINT | Digital forensics |
| Gary Klein | Naturalistic Decision Making, Recognition-Primed Decision model | Cognitive science |
| Ben Gilad | Competitive Intelligence, SCIP, strategic early warning systems | Business strategy |
| John Ioannidis | "Why Most Published Research Findings Are False", PPV, meta-research | Research integrity |
| Daniel Kahneman | System 1/System 2, cognitive biases, Noise, decision hygiene | Behavioral economics |
squads/deep-research/
├── config.yaml # Squad configuration and pipeline architecture
├── README.md # This file
├── CHANGELOG.md # Version history
├── ARCHITECTURE.md # Pipeline architecture deep dive
├── HEADLINE.md # Squad value proposition
├── agents/
│ ├── dr-orchestrator.md # Orchestrator: Pipeline Coordinator
│ ├── sackett.md # Tier 0: PICO question formulation
│ ├── booth.md # Tier 0: Methodology selection (SALSA)
│ ├── creswell.md # Tier 0: Research design (mixed methods)
│ ├── forsgren.md # Tier 1: Technical metrics (DORA/SPACE)
│ ├── cochrane.md # Tier 1: Systematic reviews (PRISMA)
│ ├── higgins.md # Tier 1: OSINT investigations
│ ├── klein.md # Tier 1: Sensemaking (NDM/RPD)
│ ├── gilad.md # Tier 1: Competitive Intelligence
│ ├── ioannidis.md # QA: Evidence reliability audit
│ └── kahneman.md # QA: Cognitive bias audit
├── tasks/ # 17 executable task definitions
├── workflows/ # Multi-phase research workflows
├── scripts/
│ ├── deep-search.ts # Search orchestration
│ └── search-providers.ts # Provider integration
└── data/
└── minds/ # Mind DNA extractions (10 files)
| Metric | Value |
|---|---|
| Total agents | 11 |
| Total lines | 5,386 |
| Avg lines/agent | 490 |
| Orchestrator | 1 (Pipeline Coordinator) |
| Tier 0 (Diagnostic) | 3 agents (sequential) |
| Tier 1 (Execution) | 5 agents (parallel) |
| QA agents | 2 (sequential, mandatory) |
| Total tasks | 17 |
| Total workflows | 3 |
| Use cases | 4 (UC-001 through UC-004) |
| Quality gates | 4 (QG-001 through QG-004) |
| Minds cloned | 10 (from 10 distinct sources) |
| Tool integrations | 3 (Exa, Context7, WebSearch) |
Deep Research Squad v1.0 -- Created by Squad Creator Pro Philosophy: "Every research question deserves a structured, evidence-based answer."