tdl（TDL 搜索）

Traceable Development Lifecycle (TDL) Skill

Overview

TDL is a template-based development process that ensures complete traceability from requirements to implementation. It supports parallel development by multiple developers or AI agents through random ID generation and distributed traceability.

Key Features

• Random Base36 IDs: 5-character IDs (e.g., a3bf2) prevent conflicts in parallel development
• Distributed Traceability: Each document maintains its own Links section - no central tracking file
• 5-Phase Workflow: Analysis → Requirements → ADR → Design → Plan & Execution
• Document Templates: Standardized templates for all document types

When to Use This Skill

Use this skill when:

• Starting a new project and want to establish traceability from the beginning
• Improving existing projects to add or enhance traceability practices
• Creating documentation (Analysis, Requirements, ADRs, Design, Plans)
• Working in parallel with other developers or AI agents
• Troubleshooting production issues and need to trace back to requirements/design
• Conducting requirement analysis and tracking implementation coverage

Interactive Mode (Claude Code only)

When running in Claude Code and the user's intent is unclear, use AskUserQuestion tool to clarify:

Document Type (if not specified in the user's request):

• Question: "What type of TDL document do you want to create?"
• Header: "Doc Type"
• Options:

- "Analysis (AN-)" - Problem exploration and discovery - "Requirement (FR/NFR-)" - Formal specification with acceptance criteria - "ADR (ADR-)" - Architecture/design decision record - "Task (T-)" - Implementation plan with design.md and plan.md

Requirement Type (when creating a requirement):

• Question: "What type of requirement is this?"
• Header: "Req Type"
• Options:

- "FR - Functional Requirement (Recommended)" - Feature or behavior specification - "NFR - Non-Functional Requirement" - Performance, security, scalability, etc.

NFR Category (when creating a non-functional requirement):

• Question: "What category does this NFR belong to?"
• Header: "Category"
• Options:

- "Performance" - Response time, throughput - "Security" - Authentication, authorization, data protection - "Scalability" - Load handling, growth capacity - "Reliability" - Availability, fault tolerance

• If user needs a different category, they can specify via the "Other" option

ADR Template (when creating an ADR):

• Question: "Which ADR template should be used?"
• Header: "Template"
• Options:

- "Full ADR (Recommended)" - Comprehensive template for significant decisions - "Lite ADR" - Simplified template for tactical decisions

TDL Workflow Phases

┌─────────────┐    ┌──────────────┐    ┌─────────┐    ┌────────┐    ┌──────────────────┐
│  Analysis   │───▶│ Requirements │───▶│   ADR   │───▶│ Design │───▶│ Plan & Execution │
│   (AN-)     │    │   (FR/NFR)   │    │ (ADR-)  │    │        │    │                  │
└─────────────┘    └──────────────┘    └─────────┘    └────────┘    └──────────────────┘
     │                    │                 │              │                  │
     │                    │                 │              │                  │
     ▼                    ▼                 ▼              ▼                  ▼
  Explore             Formalize         Document       Technical       Implementation
  problem             what/why          decisions       design           phases

Phase 1: Analysis (AN-xxxxx)

Purpose: Explore problem space and discover requirements

Output: docs/analysis/AN-xxxxx-topic.md

python scripts/create_analysis.py "User Authentication Flow"

Lifecycle: Draft → Active → Complete → Archived (after requirements formalized)

Phase 2: Requirements (FR-xxxxx / NFR-xxxxx)

Purpose: Formalize what needs to be built with measurable acceptance criteria

Output:

• Functional: docs/requirements/FR-xxxxx-topic.md
• Non-Functional: docs/requirements/NFR-xxxxx-topic.md

python scripts/create_requirement.py "User Authentication" --type FR
python scripts/create_requirement.py "API Response Time" --type NFR --category Performance

Lifecycle: Proposed → Accepted → Implemented → Verified → (Deprecated)

Phase 3: Architecture Decision Records (ADR-xxxxx)

Purpose: Document important design decisions and their rationale

Output: docs/adr/ADR-xxxxx-topic.md

python scripts/create_adr.py "Use PostgreSQL for data storage"
python scripts/create_adr.py "Minor API change" --lite  # For tactical decisions

Lifecycle: Proposed → Accepted | Rejected → (Deprecated | Superseded by ADR-xxxxx)

Phase 4 & 5: Design and Plan (T-xxxxx)

Purpose: Technical design and phased implementation plan

Output: docs/tasks/T-xxxxx-topic/

• design.md: How to implement (technical design)
• plan.md: What to do (phased implementation)

python scripts/create_task.py "implement-user-auth" --requirements FR-a1b2c,FR-d3e4f

Lifecycle: Not Started → Phase X In Progress → Blocked → Under Review → Completed

Random ID System

TDL uses 5-character Base36 random IDs instead of sequential numbers to prevent conflicts in parallel development.

Why Random IDs?

ID Format

• Characters: 0-9 and a-z (Base36)
• Length: 5 characters
• Combinations: ~60 million
• Collision probability: ~1% at 1,100 documents

Generate IDs

# Generate a raw ID
python scripts/tdl_new_id.py

# Generate with prefix
python scripts/tdl_new_id.py --prefix FR
# Output: FR-a3bf2

Directory Structure

docs/
├── analysis/           # Problem exploration (AN-xxxxx)
│   ├── archive/        # Completed analyses
│   └── AN-a3bf2-topic.md
├── requirements/       # Formal requirements (FR/NFR-xxxxx)
│   ├── FR-b4cd8-topic.md
│   └── NFR-c5de9-topic.md
├── adr/               # Architecture decisions (ADR-xxxxx)
│   ├── archive/       # Deprecated/superseded ADRs
│   └── ADR-d6ef0-topic.md
├── tasks/             # Implementation tasks (T-xxxxx)
│   └── T-e7fa1-topic/
│       ├── design.md
│       └── plan.md
└── templates/         # Document templates

Distributed Traceability

Every TDL document has a Links section that maintains relationships to other documents:

## Links

- **Analysis**: [AN-a3bf2](../analysis/AN-a3bf2-topic.md)
- **Requirements**: [FR-b4cd8](../requirements/FR-b4cd8-topic.md)
- **Related ADRs**: N/A – First implementation
- **Issue**: #123
- **PR**: N/A – Not yet submitted

Rules for Links

1. Always include all link categories - use N/A – <reason> if not applicable
2. Use relative paths for internal links
3. External references go in a separate section
4. Update links when relationships change

Available Scripts

Initialize TDL Structure

python scripts/init_tdl_docs.py [path]

Creates the complete directory structure and README files.

Generate Random ID

python scripts/tdl_new_id.py [--prefix PREFIX]

Generates a unique 5-character Base36 ID with optional prefix.

Create Documents

# Analysis
python scripts/create_analysis.py "Topic Name"

# Requirements
python scripts/create_requirement.py "Title" --type FR|NFR [--category Category]

# ADR
python scripts/create_adr.py "Decision Title" [--lite]

# Task (with design.md and plan.md)
python scripts/create_task.py "task-name" [--requirements FR-xxxxx,NFR-xxxxx]

Analyze Traceability

# Show status summary
python scripts/trace_status.py

# Detailed report
python scripts/trace_status.py --verbose

# CI mode (exit 1 if gaps found)
python scripts/trace_status.py --check

# Markdown output
python scripts/trace_status.py --format markdown

Document Templates

Common Structure

All documents follow this structure:

# Title

## Metadata

- **ID**: [TYPE-xxxxx]
- **Type**: [Document Type]
- **Owner**: [Person or role]
- **Reviewers**: [List]
- **Status**: [Status]
- **Date**: YYYY-MM-DD

## Links

- **[Link Type]**: [ID or N/A – reason]
...

## [Content Sections]
...

## External References

- [External links]

Template Rules

1. Language: All documents in English
2. Date format: YYYY-MM-DD
3. ID placement: In Metadata, not in title
4. Links section: Required in all documents

Code Integration

In Commits

git commit -m "feat(auth): Implement JWT authentication

Implements FR-b4cd8 User Authentication
Related ADR: ADR-d6ef0

Closes: #123"

In Code Comments

class AuthMiddleware:
    """JWT authentication middleware.

    Implements: FR-b4cd8
    Design: docs/tasks/T-e7fa1-implement-auth/design.md
    """

In Logs

logger.info(
    "User authenticated",
    extra={
        "requirement": "FR-b4cd8",
        "user_id": user_id,
        "action": "auth.login.success"
    }
)

Best Practices

Document Management

1. Create documents using scripts - ensures proper ID generation and structure
2. Update Links immediately - when relationships are established
3. Archive completed analyses - after requirements are formalized
4. Update status regularly - reflect current state

Parallel Development

1. Always use random IDs - never manually assign sequential numbers
2. Check traceability before merging - python scripts/trace_status.py --check
3. No central tracking file - rely on Links sections in documents
4. Resolve conflicts in content, not IDs - IDs should never conflict

Quality Assurance

1. Run traceability analysis - before releases and in CI
2. Fill all gaps - orphan requirements need tasks, orphan tasks need requirements
3. Review Links section - in code reviews

Quick Start

For New Projects

# 1. Initialize structure
python scripts/init_tdl_docs.py

# 2. Start with analysis
python scripts/create_analysis.py "Initial Project Analysis"

# 3. Create requirements from analysis
python scripts/create_requirement.py "Core Feature" --type FR

# 4. Document design decisions
python scripts/create_adr.py "Technology Stack Selection"

# 5. Create implementation task
python scripts/create_task.py "implement-core-feature" --requirements FR-xxxxx

For Existing Projects

# 1. Initialize structure (won't overwrite existing)
python scripts/init_tdl_docs.py

# 2. Check current traceability
python scripts/trace_status.py

# 3. Document existing architecture
python scripts/create_adr.py "Current Architecture Overview"

# 4. Gradually add traceability to new work

Reference Materials

The references/ directory contains:

• adr-template.md - ADR writing guide with examples
• commit-message-guide.md - Commit message standards
• pr-template.md - Pull request templates
• logging-patterns.md - Structured logging patterns
• api_reference.md - Script API documentation

Troubleshooting

"ID collision detected"

The script automatically retries up to 10 times. If this persists:

• Check if docs/ directory has an unusually large number of documents
• Manually verify the generated ID doesn't exist

"Traceability gaps identified"

Run python scripts/trace_status.py --verbose to see:

• Requirements without implementing tasks
• Tasks without linked requirements

"Links section missing"

All documents must have a Links section. Use the creation scripts to ensure proper structure.

Summary

TDL provides a systematic approach to software development traceability:

Start small, use the provided scripts, and gradually build a culture of traceability.