| name | tf-analyze |
| description | Analyze requirements and map out components, roles, and dependencies. |
| allowed-tools | Bash, Read, Glob, Grep, Write, Edit |
Marblo Requirements Analysis
The step where you analyze requirements and understand the structure before writing any code.
This skill does not create tasks. It only performs analysis.
When the analysis is done, create tasks with /tf-create-tasks.
⛔ Required rule: Marblo MCP only
Never use the Claude Code built-in tools (TaskCreate, TaskList, TaskUpdate, TaskGet).
All task-related work MUST use the Marblo MCP tools.
Phase 1: Gather requirements
Take the user's input (natural language, documents, URLs, etc.) and grasp the essentials.
Must confirm
- Goal: "What problem does this project/feature solve?"
- Users: "Who uses it?"
- Core features: "What are the 3 must-have features?"
- Tech stack: for an existing project, analyze the codebase; for a new one, recommend a stack
- Scope: MVP vs. full feature set
Phase 2: Codebase analysis (for existing projects)
When adding a feature to an existing project, analyze the codebase first.
- Understand the project structure: review the directory tree with
Glob
- Confirm the tech stack: package.json, requirements.txt, etc.
- Understand the existing patterns: API routing, DB models, component structure
- Identify the impact area: list the files/modules that need changes
Phase 3: Output the analysis result
🔍 Requirements analysis complete
━━━━━━━━━━━━━━━━━━━━
📦 Project: {project_name}
🎯 Goal: {one-line summary}
📐 Component analysis:
Backend:
- {component 1}: {description}
- {component 2}: {description}
Frontend:
- {component 1}: {description}
👤 Roles needed:
- backend: {N} tasks expected
- frontend: {N} tasks expected
- test: {N} tasks expected
🔗 Dependency graph:
Layer 1 (foundation): DB schema, project setup
Layer 2 (API): core endpoints
Layer 3 (UI): screen implementation
Layer 4 (integration): tests, deployment
⚠️ Risks/considerations:
- {risk 1}
- {risk 2}
💡 Next step: create tasks with `/tf-create-tasks`.
Notes
- Do not write code during the analysis step
- Always ask the user about anything ambiguous
- Difference from
/tf-plan: /tf-plan goes all the way to writing a PRD document; /tf-analyze only does a quick analysis