agents/openai.yaml is an extended, product-specific config intended for the machine/harness to read, not the agent. Other product-specific config can also live
Openai_Yaml centralizes product-specific configuration for Metaflow agents, providing a structured YAML format that machines and the harness read directly. It defines interface elements like display names, descriptions, icons, branding colors, and default prompts, alongside dependencies such as external tools and connection details. This setup standardizes how agents access and apply product-level settings, ensuring consistent behavior across deployments.
This skill is designed for performance marketers and growth leads who manage automated campaigns relying on consistent agent behavior and branding. It also serves SEO and PPC operators who need to integrate third-party APIs or services seamlessly within their workflows. Agencies that handle multiple client projects benefit from this skill’s ability to centralize configuration, reducing setup time and errors when scaling agent deployments.
Practitioners start by defining user-facing interface elements—such as display name, short description, and brand colors—to align the agent’s presentation with campaign branding. Next, they specify dependencies by declaring external tools, including connection types and URLs, ensuring that agents interact correctly with APIs like GitHub’s MCP server. Finally, policy settings are configured to control invocation behavior, which influences how and when the skill is triggered during runtime. This structured approach simplifies iteration and testing across multiple environments.
Can I customize the skill’s user interface elements to match different campaigns? Yes, you can adjust display names, descriptions, icons, and brand colors in the YAML to fit specific branding needs. Does the skill support multiple external tool integrations? It currently supports MCP-type tools and allows specifying multiple dependencies with detailed connection parameters. How does the invocation policy affect agent behavior? Setting `allow_implicit_invocation` to false prevents automatic injection, requiring explicit calls to the skill, which helps control resource use and invocation timing.
Attach the Openai_Yaml skill to a Metaflow agent task by referencing its path in the agent’s configuration, enabling the harness to parse and apply its settings automatically. Expect the skill to provide consistent UI elements and manage external tool connections based on the YAML definitions, streamlining deployment and maintenance. This foundational skill integrates smoothly with other Metaflow components to support complex marketing workflows and —
For broader context, see our roundup of marketing skills claude, and read how to create Claude skills for related setup guidance.