@dzhechkov/keysarium

AI Case Research Toolkit for Claude Code

Full 7-phase pipeline for AI case studies, hackathons, and casariums. Provides skills, commands, rules, shards, agent templates, and brain portability for Claude Code.

Quick Start

# One-command install via npx (no global install needed)
npx @dzhechkov/keysarium

# Or install globally
npm install -g @dzhechkov/keysarium
keysarium init

# Or without npm (standalone)
curl -fsSL https://raw.githubusercontent.com/dzhechko/product-keysarium-2026/main/packages/dz-keysarium/install.sh | bash

After installation, open Claude Code in your project directory and start using slash commands right away.

What You Get

| Component | Count | Description | |-----------|-------|-------------| | Skills | 8 | explore, frontend-design, goap-research, presentation-storyteller, problem-solver, reverse-engineering, bto, knowledge-extractor | | Commands | 24 | /casarium, /new-research, /parallel-research, /discovery, /explore-case, /research, /cjm-prototype, /solve, /architecture-phase, /presentation, /harvest, /brain-export, /brain-import, /bto, /bto-build, /bto-test, /bto-optimize, /bto-benchmark, /feature-adr, /init-platform, /learning-stats, /dream, /workers, /verify-chain | | Rules | 16 | research-quality, checkpoint-protocol, agent-swarm, domain-specific, anti-patterns, file-conventions, modular-reuse, feedback-loops, model-routing, trust-tiers, bto-quality-gates, background-workers, dream-cycles, reward-learning, witness-chain, feature-adr-conventions | | Shards | 9 | Context shards for each phase: phase-0-discovery, phase-1-explore, phase-2-research, phase-25-cjm, phase-3-solve, phase-4-architecture, phase-5-presentation, bto-evaluation, feature-adr | | Agent Templates | 5 | discovery-worker, research-worker, cjm-variant-worker, presentation-worker, parallel-research-orchestrator | | Library | 4 | phase-utils, agent-patterns, skill-loader, domain-templates | | Documentation | 12 | Deployment, admin, user guide, infrastructure, architecture, flows (for Keysarium and BTO) |

Everything is installed into your project's .claude/ directory and works natively with Claude Code.

7-Phase Pipeline

Phase 0     Phase 1     Phase 2     Phase 2.5      Phase 3     Phase 4         Phase 5
DISCOVERY   EXPLORE     RESEARCH    CJM PROTO      SOLVE       ARCHITECTURE    PRESENTATION
  15%         5%          15%         10%            15%          15%             20%

| Phase | Command | What It Does | |-------|---------|--------------| | Phase 0 -- Discovery | /discovery | Deep-dive into the product/company: JTBD analysis, competitor landscape, ROI estimation. Powered by the reverse-engineering skill. | | Phase 1 -- Explore | /explore-case | Adaptive clarification of the case brief. Asks the right questions, structures ambiguity into actionable scope. | | Phase 2 -- Research | /research | Paranoid-mode research: analogues, technologies, regulations. Every claim must have a verifiable source. | | Phase 2.5 -- CJM Prototype | /cjm-prototype | Customer Journey Map prototyping with multiple variants. Generates a working React prototype. Mandatory -- never skip this phase. | | Phase 3 -- Solve | /solve | Solution strategy using TRIZ + Game Theory. Produces process diagrams (as-is / to-be). | | Phase 4 -- Architecture | /architecture-phase | Technical architecture: C4 diagrams, sequence flows, component design, integration plan. | | Phase 5 -- Presentation | /presentation | Storytelling-driven presentation, speaker script, Q&A preparation, and executive summary. |

Each phase produces concrete artifacts (Markdown documents, Mermaid diagrams, JSX prototypes) stored in researches/<case-slug>/.

CLI Commands

npx @dzhechkov/keysarium                    # Full install (interactive, same as init)
npx @dzhechkov/keysarium init               # Install all components
npx @dzhechkov/keysarium init --minimal     # Only .claude/ (no docs, no lib)
npx @dzhechkov/keysarium init --with-docs   # Include documentation
npx @dzhechkov/keysarium init --force       # Overwrite existing files
npx @dzhechkov/keysarium init --dry-run     # Preview without making changes
npx @dzhechkov/keysarium update             # Update to latest version
npx @dzhechkov/keysarium remove             # Clean uninstall
npx @dzhechkov/keysarium list               # Show installed components
npx @dzhechkov/keysarium doctor             # Health check

Flags can be combined:

npx @dzhechkov/keysarium init --minimal --force --dry-run

Usage After Install

Full Pipeline

# Open Claude Code in your project
claude

# Start a full case study (runs all 7 phases with checkpoints)
/casarium Analyze how AI can automate customer support in banking

Individual Phases

# Product discovery
/discovery Analyze Tinkoff Bank

# Case exploration
/explore-case banking customer support automation

# Research phase
/research AI-powered document processing in insurance

# CJM prototyping
/cjm-prototype mobile banking onboarding flow

# Solution design
/solve reduce customer churn using predictive analytics

Multi-Case Research

# Run multiple cases in parallel
/parallel-research banking automation | retail personalization | healthcare triage

# Create a new isolated research
/new-research smart-warehouse-logistics

# Extract learnings after a project
/harvest researches/bank_kc_automation/
/harvest all

Example: Russian-Language Case (bilingual support)

/casarium Как AI может автоматизировать процесс обработки обращений клиентов в банке, сократив время ответа с 4 часов до 15 минут?

The toolkit processes cases in any language and adapts domain rules accordingly.

Domain Support

The toolkit automatically detects the domain from the case description and applies domain-specific rules:

Banking / FinTech

• On-premise LLM deployment (GigaChat, YandexGPT, open-source models)
• Regulatory compliance: FZ-152, Central Bank requirements, FSTEC
• Human-in-the-Loop mandatory for decision-making
• Data never leaves the security perimeter
• Visual palette: Blue / Navy / Silver

Retail / E-commerce

• Latency budget: < 200ms for real-time recommendations
• A/B testing as primary validation method
• Personalization vs. privacy balance (GDPR / FZ-152)
• Seasonality and cold-start handling
• Visual palette: Amber / Orange

Enterprise / B2B

• Change management strategy (people resist AI)
• Legacy system integration planning
• SLA and fault tolerance definitions
• ROI expressed in FTE / hours saved
• Visual palette: Teal / Indigo

Healthcare

• HITL mandatory for all clinical decisions
• Medical device regulations (FZ-323)
• AI explainability requirements
• Patient data isolation (FZ-152 + medical specifics)

Agent Swarm

The toolkit leverages Claude Code's agent capabilities for parallel execution, significantly reducing research time:

| Phase | Parallel Agents | Tasks | |-------|-----------------|-------| | Phase 0 | 2 agents | JTBD analysis || Competitors + ROI | | Phase 2 | 3 agents | Analogues || Technologies || Regulations | | Phase 2.5 | 3 agents | CJM Variant A || Variant B+C || Trend Research D | | Phase 5 | 3 agents | Presentation || Speaker Script || Q&A + Executive Summary |

Cross-case parallelism is also supported:

# Runs Phase 0 for up to 4 cases simultaneously
/parallel-research case1 | case2 | case3 | case4

Each research is fully isolated in its own researches/<slug>/ directory.

Research Artifacts

Every completed research produces a structured set of files:

researches/<case-slug>/
├── 00_product_discovery.md          # Phase 0 output
├── 01_case_brief.md                 # Phase 1 output
├── 02_research_findings.md          # Phase 2 output
├── 02.5_trend_brief.md              # Phase 2.5 output
├── 03_solution_strategy.md          # Phase 3 output
├── 04_architecture.md               # Phase 4 output
├── 05_presentation_content.md       # Phase 5 output
├── 06_speaker_script.md             # Phase 5 output
├── 07_qa_preparation.md             # Phase 5 output
├── 08_executive_summary.md          # Phase 5 output (mandatory)
├── prototype/
│   └── cjm-prototype.jsx           # Phase 2.5 React prototype
├── diagrams/
│   ├── architecture-c4.mermaid
│   ├── sequence-main-flow.mermaid
│   ├── process-as-is.mermaid
│   └── process-to-be.mermaid
└── README.md

Extending the Toolkit

Add a Custom Skill

Create a new directory under .claude/skills/:

.claude/skills/my-custom-skill/
├── SKILL.md          # Skill instructions (loaded by phases)
└── examples/         # Reference materials (optional)

Add a Custom Command

Create a new file under .claude/commands/:

.claude/commands/my-command.md

Commands follow a standard pattern: load required skills, execute logic, produce artifacts, display a checkpoint.

Add a Custom Rule

Create a new file under .claude/rules/:

.claude/rules/my-rule.md

Rules are automatically loaded by Claude Code and enforced across all phases.

Knowledge Harvesting (knowledge-extractor skill)

After completing research, extract reusable patterns using the multi-agent knowledge-extractor:

/harvest researches/<slug>/                    # Full harvest from one directory
/harvest researches/<slug>/ only patterns      # Extract only patterns
/harvest all                                   # Harvest from all researches
/harvest all only rules,templates              # All researches, specific categories
/harvest features/my-feature/                  # Harvest from a feature directory

Pipeline: Extract (5 parallel agents) → Classify (7 categories) → User Checkpoint → Gate (8 quality checks) → Integrate (auto-place)

| Stage | What Happens | |-------|-------------| | Extract | 5 parallel agents scan through different lenses: patterns, commands, rules, templates, snippets | | Classify | Findings sorted into 7 categories (skills, commands, hooks, rules, templates, patterns, snippets) with cross-dedup | | User Review | Numbered list of findings with granular control (#N commands to accept/reject/merge) | | Quality Gates | 8 blocking gates in 2 passes: deterministic (G1-G2, G5-G7) + semantic via haiku (G3-G4, G8) | | Integrate | Auto-placement into .claude/ directories + TOOLKIT_HARVEST.md update + harvest report |

The skill is domain-agnostic — works in any project, not just Keysarium research pipelines. Optional integrations (reward learning, dream cycles, brain export) activate automatically when their infrastructure is present.

Brain Portability (v1.1)

Export and import accumulated knowledge between projects:

# Export knowledge as a portable JSON container (v1.1 with manifest + checksum)
/brain-export

# Delta export — store only changes since last export (COW mode)
/brain-export --delta keysarium-brain-2026-03-01.json

# Import a brain container into another project
/brain-import path/to/brain.json

v1.1 features: SHA-256 integrity manifest, delta exports via JSON Patch (RFC 6902), 2-tier memory index, HOT/WARM/COLD/PURGE record lifecycle. Backward compatible with v1.0 brain files.

Context Shards

Shards (.claude/shards/) are lightweight context fragments that phases load on demand. They keep the main CLAUDE.md lean while providing phase-specific instructions, semantic completion promises, and quality checklists.

Requirements

• Claude Code CLI -- installed and configured (installation guide)
• Node.js >= 16.0.0 -- required for the npm install method
• Git -- required for the standalone install method

License

MIT

Links

• GitHub: https://github.com/dzhechko/product-keysarium-2026
• Issues: https://github.com/dzhechko/product-keysarium-2026/issues
• npm: https://www.npmjs.com/package/@dzhechkov/keysarium