create-agents

Reusable agent manifests require explicit model profile metadata and a structured body containing role, core contract, uncertainty protocol, and evaluation guidelines to ensure portability across shared software project libraries.

A harness contract in agent development defines the structured execution boundary, including the agent role, core contract, fallback behavior, and evaluation guidance. This establishes consistent operational parameters for template-driven execution.

Yes, meta-prompting supports template-driven agent development for bounded coding tasks. You can create a portable manifest with required frontmatter and optional references to structure execution behavior.

No specific dependencies are required for skill authoring. The manifest enforces a frontmatter with name and description, while optional components like references and assets organize shared knowledge without external tooling constraints.

The best way to structure an agent manifest is using a frontmatter with name and description, followed by a structured body with role, core contract, uncertainty protocol, and evaluation guidelines, plus optional references and assets.

Add evaluation guidelines to an agent manifest when establishing team conventions for testing and shared agent directories. They provide structured execution guidance and fallback behavior for consistent software project outcomes.