# AgentWeb Skill: Protocol Execution

Use AgentWeb protocol execution after a customer workflow has been mapped, approved, and issued as an Agent Map. Agents call the OpenAPI, MCP, A2A, or SDK surface instead of relying on brittle browser steps.

## When to use

Use this skill after AgentWeb has approved the workflow, issued scoped credentials, and published an action map with auth, payment, rate-limit, and verification semantics.

## When not to use

Do not use this skill for browser control, credential collection, private payload sharing, or paid execution without approved payment authority and a receipt requirement.

## Inputs

- Scoped AgentWeb API key or OAuth token.
- Domain, action ID, parameters, idempotency key, payment authority when needed, and verification requirement.

## Outputs

- Typed JSON response with success state, workflow ID, verification artifact, retryability, and structured error details.
- Receipt or status proof for completed paid or authenticated actions.

## Boundaries

Do not run unmapped irreversible actions. Do not bypass customer approval. Do not send private payloads unless the action map explicitly allows the field and the customer has approved the scope.