AI Skills

AI Skills are the mechanism that makes OpenLogos executable — not just a set of principles you have to remember, but operational instructions that AI agents read and follow automatically.

What is a Skill?

A Skill is a Markdown file (SKILL.md) stored in the project’s skills/ directory. Each Skill corresponds to a specific phase or task in the OpenLogos workflow. When a developer asks an AI tool to perform a task, the tool reads the relevant Skill and follows its structured instructions.

Key properties:

Platform-agnostic: Works with Claude Code, OpenCode, Cursor, or any tool that reads project files
Phase-aware: Each Skill knows which phase it belongs to and what inputs it expects
Guardrailed: Skills enforce methodology rules (e.g., “do not generate code without test cases”)
Self-contained: Each Skill includes its own instructions, output templates, and quality checks

Built-in Skills

OpenLogos includes 13 Skills covering the entire lifecycle:

Phase 1 — WHY

Skill	Purpose
`project-init`	Initialize project structure and configuration
`prd-writer`	Write product requirements with scenario-driven acceptance criteria

Phase 2 — WHAT

Skill	Purpose
`product-designer`	Create feature specs, interaction flows, and HTML prototypes

Phase 3 — HOW

Skill	Purpose
`architecture-designer`	Design technical architecture and select tech stack
`scenario-architect`	Model scenarios as sequence diagrams with API calls
`api-designer`	Design OpenAPI specifications derived from sequence diagrams
`db-designer`	Design database schema from API and scenario requirements
`test-writer`	Write unit + scenario test case documents (before code)
`test-orchestrator`	Design API orchestration tests (API projects only)
`code-implementor`	Generate business code and test code with spec fidelity
`code-reviewer`	Review code against the full specification chain

Cross-phase

Skill	Purpose
`change-writer`	Write change proposals with impact analysis
`merge-executor`	Merge delta files into main documents

How Skills are loaded

Developer asks AI to perform a task (e.g., “write the requirements”)
AI reads AGENTS.md → detects current phase → identifies relevant Skill
AI reads the SKILL.md file for operational instructions
AI follows the Skill’s step-by-step process, using the Skill’s output templates and quality checks

The loading mechanism varies by platform:

Platform	Mechanism
Claude Code	Native `.claude/` plugin system
OpenCode	`hooks/` and command integration
Cursor	`.cursor/rules/` auto-attached rules that reference Skills
Other tools	`AGENTS.md` as the universal entry point

Skill anatomy

A typical SKILL.md contains:

# Skill Name

## Trigger
When to activate this Skill.

## Inputs
What documents/artifacts this Skill reads.

## Process
Step-by-step instructions the AI follows.

## Output Template
The expected format of the output document.

## Quality Checks
Self-verification before marking the task complete.

Custom Skills

You can create project-specific Skills by adding SKILL.md files to the skills/ directory and referencing them in logos-project.yaml. Custom Skills follow the same format as built-in ones.

See also: Full Skills Reference for detailed documentation of each Skill →