Documentation

Action Types Reference

Action types classify an action; they do not show whether Open for Agents can execute it. Common labels include search, contact_form, lead_form, add_to_cart, checkout_start, and auth.

Detection and action classification do not prove that an action can be executed. Form providers have separate tested-support, default-enablement, field, and version boundaries in the Integrations matrix. Tested form submission currently covers Contact Form 7 and WPForms Lite.

Detected Action Types

Action Type	Description
`search`	Searches public site content without changing it.
`contact_form`	Detected contact workflows, including reviewed Contact Form 7 forms.
`lead_form`	Detected lead workflows, including reviewed WPForms Lite forms.
`browse_products`	WooCommerce catalog/shop browsing actions.
`add_to_cart`	Adds products to cart.
`view_cart`	Views current cart state.
`update_cart`	Updates quantity/coupon/cart state.
`checkout_start`	Classifies a detected checkout entry point; Open for Agents does not provide a tool to execute it.
`order_status`	Identifies an authenticated order or account action; supported agents cannot execute it.
`auth`	Identifies login, registration, or password-reset actions; supported agents cannot execute them.
`submit_comment`	WordPress comment submission actions.
`custom`	Other inferred actions, including detected AJAX and REST endpoints.

Risk Levels

Low: Read-oriented or low-impact interactions such as search.
Medium: State-changing interactions with limited commercial impact, typically non-payment forms.
High: Sensitive operations such as authentication, checkout, or account-management paths.

Form submission controls

Eligible form tools are irreversible and disabled by default. The exact provider form, supported field schema, notifications, redirects, and configured feeds require administrator review. The visitor then sees the extracted values and explicitly approves or cancels before the provider performs its own validation and submission.

Confidence Scale

High confidence (>= 0.80): Detector matched a known plugin API or strongly typed integration pattern.
Medium confidence (>= 0.50 and < 0.80): Action inferred from structured form semantics and endpoint behavior.
Low confidence (< 0.50): Action discovered through generic HTML or hook/route mining and should be manually reviewed before publication.

Auth Required and Restricted Flags

Auth required indicates that the action depends on a logged-in session or privileged capability. Restricted means the action is intentionally withheld from public tiers unless you explicitly allow it.

Admin UI Glossary

UI Label	Meaning	Publication Impact
Risk	Operational impact level: low, medium, high.	Dangerous/high-risk actions are excluded from public tiers.
Confidence	How certain the detector is about action semantics.	Use low confidence as a manual-review signal before publishing.
Auth Required	Action depends on authenticated session or privileged context.	Auth-required actions are excluded from `public_standard`.
Restricted	A site owner has chosen to keep this action private.	Restricted actions are treated as dangerous and stay out of public tiers.
Status	Enabled or disabled in the local action map.	Disabled actions are omitted from all exports.
Agent Visibility	Allowed agent contexts for discovering an action.	Sensitive extension-visible actions are filtered in public exports.

Next: Publication Tiers