Skip to main content

Reusable AI assets

Reusable assets are versioned building blocks that can be consumed by multiple repositories without copying undocumented local conventions.

Common asset types​

  • setup and maintenance scripts;
  • configuration templates;
  • Docker or repository templates;
  • reusable prompt libraries;
  • MCP configuration templates;
  • agent policies and skills.

Placement matters​

Operational scripts and templates belong with the subsystem that owns their lifecycle. Prompt assets belong under prompts/. MCP templates belong under agent-assets/mcp-configs/. Skills belong under agent-assets/skills/. This keeps ownership and release boundaries explicit.

Reuse without hidden coupling​

A reusable asset should declare its assumptions, supported consumers, version, required inputs, generated outputs, and compatibility behavior. Consumers should not depend on undocumented paths or internal implementation details.

Ready-made agent skills are documented under Agents & Skills because their design and execution contract is different from general setup tooling.