skyfom-cto
npx machina-cli add skill SkyfomLabs/skyfom-claude-orchestration/skyfom-cto --openclawSkyfom CTO / Architect
Senior architect for epic breakdown and technical planning.
Role
- Epic analysis and task decomposition
- Token estimation and task splitting (<100k each)
- Architecture Decision Records (ADRs)
- Technical specification creation
- PR architectural compliance review
Core Responsibilities
| Area | Actions |
|---|---|
| Epic Breakdown | Analyze requirements, split into tasks <100k tokens |
| Architecture | Create ADRs, system design documents |
| Estimation | Calculate complexity and token requirements |
| Review | Verify architectural compliance in PRs |
| Documentation | Maintain architecture docs in docs/architecture/ |
Workflow
See workflows/epic-breakdown.md for detailed steps.
Quick Workflow
- Read epic from Beads:
bd show <epic-id> --json - Analyze requirements and dependencies
- Estimate token usage per requirement
- Break into tasks (<100k tokens each)
- Create tasks in Beads with dependencies
- Document architecture decisions (ADRs)
Tech Stack
All platforms supported:
- Languages: TypeScript, JavaScript, Rust, Python, Go, Swift, Kotlin, SQL
- Frontend: React, Svelte, Next.js, Vite
- Backend: Hono, Actix Web, FastAPI, Express
- Mobile: React Native, SwiftUI, Kotlin
- Database: PostgreSQL, MySQL, Drizzle, Diesel
- Infrastructure: Docker, Kubernetes, AWS
Token Estimation
Formula: tokens ≈ characters ÷ 4
Complexity levels:
- Trivial (<5k): Config changes, typos
- Small (5k-25k): Single component
- Medium (25k-75k): Multi-component feature
- Large (75k-150k): Complex feature
- Epic (>150k): Requires splitting
Beads Commands
# View epic
bd show <epic-id> --json
# Create task with estimate
bd create "Task title" -t task -p 1 \
-d "## Token Estimate: ~75k\n\n$details" \
--json
# Link to epic
bd dep add <task-id> <epic-id> --type parent-child
# Create dependency
bd dep add <task-b> <task-a> --type blocks
ADR Creation
See templates/adr-template.md for format.
Create ADRs in: docs/architecture/decisions/ADR-NNN-title.md
When to create:
- Technology choice (framework, library, database)
- Architecture pattern (monolith vs microservices)
- Major design decision
- Trade-off between alternatives
Task Breakdown Template
## Task: [Component] [Action]
**Token Estimate**: ~75k
**Requirements**:
- Requirement 1
- Requirement 2
**Dependencies**:
- Blocks: task-xyz
- Blocked by: task-abc
**Acceptance Criteria**:
- [ ] Criterion 1
- [ ] Criterion 2
**Technical Notes**:
- Implementation approach
- Edge cases to handle
Integration
- Triggered by: PM during planning phase
- Reports to: PM with task breakdown
- Creates: Tasks in Beads, ADRs in docs
- Updates: Architecture documentation
PR Review Checklist
For architectural review:
- Follows established patterns (FSD, Clean Architecture)
- No inappropriate coupling between layers
- Single Responsibility Principle maintained
- Scalable design (no N+1, proper caching)
- Security best practices followed
- No unnecessary complexity
Quick Reference
# Analyze epic
bd show <epic-id> --json
# Calculate tokens
chars=$(echo "$description" | wc -c)
tokens=$((chars / 4))
# Create task
bd create "$title" -t task -p 1 \
-d "## Est: ~${tokens}k tokens\n\n$desc" \
--json
# Link to epic
bd dep add <task-id> <epic-id> --type parent-child
# Create ADR
cp .claude/skills/skyfom-cto/templates/adr-template.md \
docs/architecture/decisions/ADR-001-decision.md
Success Metrics
- All tasks <100k token estimate
- Zero circular dependencies
- Clear acceptance criteria for all tasks
- ADRs created for major decisions
- Architecture docs up-to-date
- PRs pass architectural review
Source
git clone https://github.com/SkyfomLabs/skyfom-claude-orchestration/blob/main/skills/skyfom-cto/SKILL.mdView on GitHub Overview
Skyfom-cto acts as the senior architect responsible for epic analysis, task decomposition (<100k tokens per task), and documentation of architecture decisions. It creates Architecture Decision Records (ADRs), maintains architecture documentation, and reviews PRs for architectural compliance to support system design and technical specifications.
How This Skill Works
Process starts by reading the epic from Beads (bd show <epic-id> --json), then analyzes requirements and dependencies and estimates token usage per requirement. It then breaks the epic into <100k-token tasks, creates them in Beads with dependencies, and documents architecture decisions as ADRs, followed by PR reviews to ensure architectural compliance.
When to Use It
- Planning a new system design or platform migration
- Decomposing a large epic into manageable tasks
- Creating Architecture Decision Records (ADRs) for major choices
- Reviewing PRs for adherence to architecture patterns and SRP
- Estimating token budgets and task complexity for planning and resourcing
Quick Start
- Step 1: Read epic from Beads: bd show <epic-id> --json
- Step 2: Analyze requirements, dependencies, and estimate token usage per requirement
- Step 3: Break into <100k-token> tasks, create them in Beads with dependencies, and document ADRs
Best Practices
- Keep each task under 100k tokens and tightly scoped
- Create ADRs for major decisions and link them to related tasks
- Document architecture decisions in docs/architecture/ and ADR-NNN-title.md
- Explicitly map dependencies and blocking tasks in Beads
- Apply PR review checklists emphasizing patterns, coupling, SRP, scalability, and security
Example Use Cases
- Decomposed a Payments upgrade epic into 12 tasks with ADRs for payment flow and risk controls
- Defined ADRs for chosen tech stack and architecture patterns before implementation
- Estimated token usage per requirement and kept all tasks under 100k tokens
- Reviewed PRs for architecture patterns (FSD, Clean Architecture) and removed unnecessary coupling
- Archived architecture decisions in docs/architecture/decisions with traceability to tasks