Templates Library
Relevant source files
The following files were used as context for generating this wiki page:
The Templates Library within the eng-platform repository serves as a centralized repository of standardized engineering configurations, documentation formats, and automated workflow patterns. Its primary purpose is to provide a "foundation that projects build on," ensuring unified engineering standards across all company projects by offering reusable assets for AI context, continuous integration, and development workflows.
Sources: README.md:3-8, CONTRIBUTING.md:7-9
The library facilitates the rapid initialization of new projects and the standardization of existing ones through a multi-tiered approach:
- AI Context Management: Standardizing how LLMs interact with codebases via
CLAUDE.md. - Workflow Standardization: Defining Pull Request structures and Git behaviors.
- Automation & CI: Providing ready-to-use GitHub Action workflows and linting/testing gates.
- Project Governance: Establishing "Protected Code" policies and documentation standards like "Chain of Thought."
Sources: README.md:12-25, internal/agents/documentation-agent-context.md:16-30
AI Context Templates
The cornerstone of the library's AI-native approach is the CLAUDE.md template. This file acts as a content-addressable context system (Pattern-CONTEXT-002) designed to reduce LLM context bloat while providing essential architectural insights and safety guardrails.
Core Components of AI Context
The template defines a structured layout for project-specific information, including:
- Quick Pattern Index: Shorthand references for architecture and data models.
- Investigation Mode Triggers: Specific phrases that invoke automated search workflows.
- Protection Status: Lists of stable components that require pre-flight checklists before modification.
- Bug Prevention: Documented anti-patterns (e.g., BUG-001) with root causes and prevention checklists.
Sources: templates/CLAUDE.md:1-124, README.md:120-123
This diagram illustrates the hierarchical structure of the AI context template used to guide LLM agents.
Sources: templates/CLAUDE.md:8-100
Project Initialization & Structure
The library provides automated skills for setting up the ÆtherLight development environment. This process involves detecting the project framework and deploying a standardized folder hierarchy.
Standardized Directory Layout
When a project is initialized, the following structure is enforced to support automated workflows:
| Directory | Purpose |
|---|---|
.claude/ | AI configuration and skill definitions |
.aetherlight/ | Pattern library and metadata |
sprints/ | TOML-based sprint definitions and task tracking |
docs/ | Architecture and pattern documentation |
analysis/ | Automated code quality reports |
Sources: skills/initialize/SKILL.md:65-76, skills/initialize/initialize-eng-platform.md:204-213
Initialization Logic
The initialization skill performs framework detection (Node.js, Rust, Python, etc.) and build system identification to tailor the template deployment.
The initialization flow ensures that all necessary templates and configurations are applied based on the detected project environment.
Sources: skills/initialize/SKILL.md:46-110
Documentation and Standards Templates
The library enforces high-quality technical writing through specific documentation patterns, most notably the Chain of Thought (CoT) standard and the standardized PR template.
Chain of Thought Standard
Every class or module is expected to follow a specific CoT template to ensure reasoning is preserved:
- Design Decision: The approach taken.
- Why: The reasoning behind the decision.
- Reasoning Chain: A numbered, step-by-step thought process.
- Pattern: Reference to existing platform patterns.
- Performance: Metrics and targets.
Sources: internal/agents/documentation-agent-context.md:32-42, internal/agents/documentation-agent-context.md:231-244
Pull Request (PR) Template
The library includes a PR_TEMPLATE.md used to enforce consistency during code reviews.
| Section | Requirements |
|---|---|
| Type of Change | Selection of Bug fix, New feature, Breaking change, etc. |
| Test Plan | Requirements for unit, integration, and manual testing results. |
| Checklist | Validation of style guidelines, documentation updates, and "Protected Code" checks. |
| Breaking Changes | Description of breaks and required migration paths. |
Sources: templates/github/PR_TEMPLATE.md:5-52
Code Protection Templates
A specialized system of annotations and policy templates is provided to prevent regressions in stable features. This system uses levels to define how code can be modified.
Protection Levels
The protect skill utilizes the following templates for code annotation:
- @immutable: Reserved for core APIs and external integrations. Interface changes are strictly prohibited.
- @protected: Applied to passing features; allows refactoring but prohibits behavior changes.
- @maintainable: Core functionality works, but bug fixes and edge-case modifications are expected.
Sources: skills/protect/SKILL.md:15-30, skills/protect/SKILL.md:73-118
The protection workflow automates the stabilization of features after testing.
Sources: skills/protect/SKILL.md:50-150
Summary
The Templates Library is the foundational layer of the eng-platform, providing the structural and procedural consistency required for autonomous agent operations and multi-project standardization. By combining AI context files (CLAUDE.md), reasoning standards (CoT), and automated initialization skills, the library ensures that every project within the ecosystem adheres to high-velocity, high-quality engineering practices while maintaining a clear audit trail of architectural decisions and code protection.