skill-architect

Designing reliable agent skills requires disciplined discovery and architecture to define clear triggers, boundaries, and workflow steps before writing the actual SKILL.md file. This ensures consistent automation and prevents mis-triggering during execution.

Progressive disclosure in skill design maps content across YAML frontmatter, the SKILL.md body, and linked files. This architecture minimizes instruction scope and ensures the resulting skill is efficient to load and deterministic in its agent workflows.

Creating an MCP-enhanced workflow skill involves outcome-first discovery to identify target tools, followed by architecture planning to map content to progressive disclosure. This process ensures the skill is composable and avoids subagent guidance.

Validating YAML frontmatter requires enforcing structural rules, trigger boundaries, and instruction clarity before delivery. This quality validation checks that required fields like name and description are present to ensure the skill is compact and deterministic.

Standalone skill architecture is not suitable when your goal is to produce technical-design documents or subagent guidance rather than reliable automation. The architecture process specifically limits instruction scope to avoid these improper use cases.

You do not need predefined components or dependencies to start building a skill. The architecture process guides you through defining a minimal folder structure and drafting trigger-focused descriptions from scratch using only your target workflow.