@dzhechkov/keysarium 1.0.0

package/templates/lib/agent-patterns.md

@@ -0,0 +1,115 @@
+# Agent Spawning Patterns — Reusable Templates
+
+## Pattern 1: Fan-Out + Merge
+
+Use when: Multiple independent research directions need to run in parallel.
+
+```
+┌─────────┐
+│ Orchestr│──┬── Agent 1 (Direction A) ──┐
+│   ator  │  ├── Agent 2 (Direction B) ──┤── Merge → Artifact
+│         │  └── Agent 3 (Direction C) ──┘
+└─────────┘
+```
+
+Implementation:
+```python
+# Pseudocode for Agent tool calls
+# Launch ALL agents in a SINGLE message (parallel execution)
+
+Agent(description="Direction A research", ...)  # concurrent
+Agent(description="Direction B research", ...)  # concurrent
+Agent(description="Direction C research", ...)  # concurrent
+
+# After all complete → synthesize into single document
+```
+
+Used in: Phase 0, Phase 2, Phase 2.5, Phase 5
+
+## Pattern 2: Pipeline (Sequential with Gates)
+
+Use when: Each step depends on the previous step's output.
+
+```
+Phase 0 → Checkpoint → Phase 1 → Checkpoint → Phase 2 → ...
+```
+
+Implementation:
+- Execute phase
+- Create artifact
+- Display checkpoint
+- WAIT for user confirmation
+- Read previous artifacts
+- Execute next phase
+
+Used in: Full /casarium pipeline
+
+## Pattern 3: Isolated Parallel (Multi-Research)
+
+Use when: Multiple completely independent research projects.
+
+```
+┌──────────┐
+│ Launcher │──┬── Agent: Case 1 (isolated dir) ──→ researches/case1/
+│          │  ├── Agent: Case 2 (isolated dir) ──→ researches/case2/
+│          │  └── Agent: Case 3 (isolated dir) ──→ researches/case3/
+└──────────┘
+```
+
+Implementation:
+- Each agent gets its own directory
+- No shared state between agents
+- Summary table after all complete
+
+Used in: /parallel-research
+
+## Pattern 4: Skill-Loaded Worker
+
+Use when: An agent needs to operate with a specific skill's methodology.
+
+```
+Agent(
+  prompt="""
+    Step 1: Read .claude/skills/{skill}/SKILL.md
+    Step 2: Read context files (previous phases)
+    Step 3: Apply skill methodology
+    Step 4: Generate artifact
+  """
+)
+```
+
+Key rules:
+- ALWAYS load skill BEFORE processing
+- ALWAYS read prior artifacts for context
+- Skill provides methodology, context provides data
+
+## Pattern 5: Background Agent
+
+Use when: Long-running task that doesn't block user interaction.
+
+```
+Agent(
+  run_in_background=True,
+  description="Background harvest analysis",
+  ...
+)
+```
+
+Used in: /harvest (reviewing large project directories)
+
+## Cost Optimization Guide
+
+| Task Complexity | Recommended Model | Token Budget |
+|----------------|-------------------|--------------|
+| File operations, formatting | haiku | ~10K |
+| Research synthesis, analysis | sonnet | ~50K |
+| Creative (CJM, presentation) | opus | ~100K |
+| Full pipeline orchestration | opus | ~500K |
+
+## Error Handling
+
+If an agent fails:
+1. Log the error
+2. Do NOT retry automatically (may waste tokens)
+3. Report to orchestrator with error context
+4. Orchestrator decides: retry, skip, or ask user

package/templates/lib/domain-templates.md

@@ -0,0 +1,172 @@
+# Domain Templates — Reusable Domain Configurations
+
+## Purpose
+Pre-configured settings for common case study domains.
+Auto-detected from case description keywords.
+
+## Template: Banking / FinTech
+
+```yaml
+domain: banking
+palette: [Blue, Navy, Silver]
+tone: strict, reliable, conservative
+emojis: [🏦, 💳, 🔒]
+
+regulatory:
+  - ФЗ-152 (Персональные данные)
+  - Требования ЦБ РФ
+  - ФСТЭК (Защита информации)
+  - PCI DSS (платёжные данные)
+
+architecture_constraints:
+  - deployment: on-premise ONLY
+  - llm_options: [GigaChat, YandexGPT, LLaMA, Mistral, Qwen]
+  - data_perimeter: closed contour
+  - hitl: MANDATORY for all decisions
+  - audit: full logging of AI actions required
+
+presentation_focus:
+  - Security and compliance
+  - Data never leaves perimeter
+  - Regulatory adherence
+  - Conservative ROI estimates
+
+typical_metrics:
+  - Время обработки заявки: часы → минуты
+  - Стоимость обработки: руб./заявка
+  - Точность классификации: %
+  - FTE экономия: сотрудников
+```
+
+## Template: Retail / E-commerce
+
+```yaml
+domain: retail
+palette: [Amber, Orange, Warm Gray]
+tone: energetic, data-driven, customer-centric
+emojis: [🛍️, 📦, ⚡]
+
+regulatory:
+  - ФЗ-152 (Персональные данные)
+  - ФЗ о рекламе
+  - GDPR (если международный)
+
+architecture_constraints:
+  - latency: < 200ms for real-time
+  - deployment: cloud OK (AWS, GCP, Yandex Cloud)
+  - ab_testing: mandatory validation method
+  - personalization_vs_privacy: explicit balance required
+
+presentation_focus:
+  - Customer experience improvement
+  - Conversion rate impact
+  - A/B test results
+  - Scalability for peak loads (Black Friday)
+
+typical_metrics:
+  - Конверсия: % увеличение
+  - Средний чек: руб. увеличение
+  - Возврат клиентов: retention %
+  - Время поиска товара: секунды → секунды
+```
+
+## Template: Enterprise / B2B
+
+```yaml
+domain: enterprise
+palette: [Teal, Indigo, White]
+tone: professional, ROI-focused, structured
+emojis: [👥, 📊, 🎯]
+
+regulatory:
+  - Корпоративные политики ИБ
+  - SLA requirements
+  - Compliance отдела
+
+architecture_constraints:
+  - deployment: hybrid (on-prem + cloud)
+  - integration: legacy systems (SAP, 1C, Bitrix)
+  - auth: LDAP/AD integration
+  - change_management: phased rollout required
+
+presentation_focus:
+  - ROI in FTE / hours saved
+  - Change management plan
+  - Legacy integration approach
+  - Phased rollout with clear criteria
+
+typical_metrics:
+  - FTE экономия: человеко-часов/мес
+  - Время процесса: дней → часов
+  - Ошибки: % снижение
+  - Employee satisfaction: NPS change
+```
+
+## Template: Healthcare
+
+```yaml
+domain: healthcare
+palette: [Green, White, Light Blue]
+tone: careful, evidence-based, patient-centric
+emojis: [🏥, 💊, 🩺]
+
+regulatory:
+  - ФЗ-323 (Охрана здоровья)
+  - ФЗ-152 (Персональные данные)
+  - Медицинские изделия (если applicable)
+  - Клинические рекомендации
+
+architecture_constraints:
+  - hitl: MANDATORY for ALL clinical decisions
+  - explainability: required for AI recommendations
+  - data_isolation: patient data encryption at rest and in transit
+  - deployment: on-premise preferred
+  - audit: complete decision trail
+
+presentation_focus:
+  - Patient safety first
+  - Clinical validation pathway
+  - Explainability of AI decisions
+  - Doctor as final decision maker
+
+typical_metrics:
+  - Точность диагностики: %
+  - Время до диагноза: часы → минуты
+  - Пропущенные случаи: % снижение
+  - Нагрузка на врача: пациентов/смену
+```
+
+## Template: Legal
+
+```yaml
+domain: legal
+palette: [Slate, Navy, Gold]
+tone: precise, authoritative, structured
+emojis: [⚖️, 📋, ✍️]
+
+regulatory:
+  - Адвокатская тайна
+  - ФЗ-152
+  - Процессуальные кодексы
+
+architecture_constraints:
+  - confidentiality: absolute
+  - hitl: MANDATORY
+  - deployment: on-premise preferred
+  - versioning: full document versioning required
+
+typical_metrics:
+  - Время подготовки документа: часы → минуты
+  - Стоимость юридического часа: руб.
+  - Покрытие нормативной базы: %
+  - Ошибки в документах: % снижение
+```
+
+## Usage
+
+In any phase, detect domain from case description and apply the corresponding template:
+- Use palette for CJM prototype design (Phase 2.5)
+- Use regulatory list for Architecture (Phase 4)
+- Use typical_metrics for Solution Strategy (Phase 3)
+- Use presentation_focus for Presentation (Phase 5)
+- Use architecture_constraints throughout

package/templates/lib/phase-utils.md

@@ -0,0 +1,107 @@
+# Phase Utilities — Reusable Components
+
+## Checkpoint Template
+
+Use this template after every phase completion:
+
+```markdown
+═══════════════════════════════════════════════════════
+⏸️ CHECKPOINT {N}: {PHASE_NAME} Complete
+{SUMMARY_LINE_1}
+{SUMMARY_LINE_2}
+Files: {FILE_LIST} ✅
+• "ок" — {NEXT_PHASE}
+• "углуби [раздел]" — доработать
+• "[feedback]" — скорректировать
+═══════════════════════════════════════════════════════
+```
+
+Variables:
+- `{N}` — Phase number (0, 1, 2, 2.5, 3, 4, 5)
+- `{PHASE_NAME}` — Human-readable phase name
+- `{SUMMARY_LINE_1}` — Key metrics/findings (1 line)
+- `{SUMMARY_LINE_2}` — Key deliverable summary (1 line)
+- `{FILE_LIST}` — Comma-separated artifact filenames
+- `{NEXT_PHASE}` — Name of next phase
+
+## Artifact Tracker Template
+
+```markdown
+ARTIFACTS (researches/{SLUG}/):
+[{STATUS}] 00_product_discovery.md     ← Phase 0
+[{STATUS}] 01_case_brief.md            ← Phase 1
+[{STATUS}] 02_research_findings.md     ← Phase 2
+[{STATUS}] 02.5_trend_brief.md         ← Phase 2.5
+[{STATUS}] 03_solution_strategy.md     ← Phase 3
+[{STATUS}] 04_architecture.md          ← Phase 4
+[{STATUS}] 05_presentation_content.md  ← Phase 5
+[{STATUS}] 06_speaker_script.md        ← Phase 5
+[{STATUS}] 07_qa_preparation.md        ← Phase 5
+[{STATUS}] 08_executive_summary.md     ← Phase 5
+[{STATUS}] prototype/cjm-prototype.jsx ← Phase 2.5
+[{STATUS}] diagrams/*.mermaid          ← Phase 3-4
+[{STATUS}] README.md                   ← Phase 6
+```
+
+Status: `x` = done, ` ` = pending
+
+## Phase Timing Calculator
+
+Given total hours `T`:
+
+| Phase | % | Minutes (T=4h) | Minutes (T=2h) | Minutes (T=6h) |
+|-------|---|-----------------|-----------------|-----------------|
+| Phase 0 | 15% | 36 | 18 | 54 |
+| Phase 1 | 5% | 12 | 6 | 18 |
+| Phase 2 | 15% | 36 | 18 | 54 |
+| Phase 2.5 | 10% | 24 | 12 | 36 |
+| Phase 3 | 15% | 36 | 18 | 54 |
+| Phase 4 | 15% | 36 | 18 | 54 |
+| Phase 5 | 20% | 48 | 24 | 72 |
+| Buffer | 5% | 12 | 6 | 18 |
+
+## Research Directory Initializer
+
+Standard structure for new research:
+```
+researches/{slug}/
+├── prototype/
+├── diagrams/
+└── README.md
+```
+
+README.md template:
+```markdown
+# Исследование: {CASE_NAME}
+
+**Создано:** {DATE}
+**Статус:** 🟡 В работе
+**Домен:** {DOMAIN}
+**Тайминг:** {HOURS} часов
+
+## Артефакты
+- [ ] Phase 0: Product Discovery
+- [ ] Phase 1: Case Brief
+- [ ] Phase 2: Research Findings
+- [ ] Phase 2.5: CJM Prototype ⭐
+- [ ] Phase 3: Solution Strategy
+- [ ] Phase 4: Architecture
+- [ ] Phase 5: Presentation
+- [ ] Phase 6: Packaging
+
+## Описание кейса
+{CASE_DESCRIPTION}
+```
+
+## Domain Auto-Detection
+
+Keywords → Domain mapping:
+
+| Keywords | Domain | Palette |
+|----------|--------|---------|
+| банк, кредит, ЦБ, платёж, финтех | Banking | Blue/Navy/Silver |
+| магазин, ритейл, товар, корзина, e-commerce | Retail | Amber/Orange |
+| HR, найм, onboarding, сотрудник, кадры | Enterprise/HR | Teal/Indigo |
+| врач, пациент, клиника, диагноз, медицина | Healthcare | Green/White |
+| суд, договор, юрист, иск, право | Legal | Slate/Navy |
+| процесс, BPM, автоматизация, workflow | Process/BPM | Violet/Gray |

package/templates/lib/skill-loader.md

@@ -0,0 +1,78 @@
+# Skill Loader — Universal Skill Loading Protocol
+
+## Purpose
+Standardized protocol for loading skills before each phase.
+Ensures consistent skill activation across Claude Code and claude.ai environments.
+
+## Loading Protocol
+
+### In Claude Code (this repository)
+```
+Read(".claude/skills/{skill-name}/SKILL.md")
+```
+
+### In claude.ai (web)
+```
+view("/mnt/skills/user/{skill-name}/SKILL.md")
+```
+
+## Skill Registry
+
+| Skill ID | Path (Claude Code) | Path (claude.ai) | Used By Phases |
+|----------|-------------------|-------------------|----------------|
+| explore | .claude/skills/explore/ | /mnt/skills/user/explore/ | Phase 1 |
+| frontend-design | .claude/skills/frontend-design/ | /mnt/skills/public/frontend-design/ | Phase 2.5 |
+| goap-research-ed25519 | .claude/skills/goap-research-ed25519/ | /mnt/skills/user/goap-research-ed25519/ | Phase 2, 2.5 |
+| presentation-storyteller | .claude/skills/presentation-storyteller/ | /mnt/skills/user/presentation-storyteller/ | Phase 5 |
+| problem-solver-enhanced | .claude/skills/problem-solver-enhanced/ | /mnt/skills/user/problem-solver-enhanced/ | Phase 3 |
+| reverse-engineering-unicorn | .claude/skills/reverse-engineering-unicorn/ | /mnt/skills/user/reverse-engineering-unicorn/ | Phase 0, 2.5 |
+
+## Phase → Skill Mapping
+
+```
+Phase 0: DISCOVERY
+  → Load: reverse-engineering-unicorn/SKILL.md
+  → Also read: modules/02-product-customers.md
+
+Phase 1: EXPLORE
+  → Load: explore/SKILL.md
+  → Also read: references/questioning-techniques.md
+
+Phase 2: RESEARCH
+  → Load: goap-research-ed25519/SKILL.md
+  → Also read: references/research-actions.md, references/source-evaluation.md
+
+Phase 2.5: CJM PROTOTYPE
+  → Load: reverse-engineering-unicorn/SKILL.md (module M2.5)
+  → Also read: modules/025-cjm-prototype.md
+  → Load: frontend-design/SKILL.md
+  → For Variant D trends:
+    → Load: goap-research-ed25519/SKILL.md
+
+Phase 3: SOLVE
+  → Load: problem-solver-enhanced/SKILL.md
+
+Phase 4: ARCHITECTURE
+  → No external skill (built-in templates)
+  → Reference: lib/phase-utils.md for domain detection
+
+Phase 5: PRESENTATION
+  → Load: presentation-storyteller/SKILL.md
+  → Also read: references/storytelling-frameworks.md, references/speaker-script-patterns.md
+```
+
+## Adding a New Skill
+
+1. Create directory: `.claude/skills/{skill-name}/`
+2. Create `SKILL.md` with YAML frontmatter:
+   ```yaml
+   ---
+   name: skill-name
+   description: What this skill does and when to trigger it.
+   ---
+   ```
+3. Add reference materials in `references/` subdirectory
+4. Add examples in `examples/` subdirectory
+5. Update this registry
+6. Update CLAUDE.md skill graph
+7. Create or update the command that uses this skill