Read what your rulesadmit from a visiting agent.

It grades your rules for AI agents that come to write and transact on your site, row by row as agent and action (browse, quote, book, buy, change, cancel). For each row you mark five layers 0 to 2: Identity, Intent, Scope, Limit and Refusal. It returns ADMIT, REVIEW or REFUSE per row and one roll-up verdict. It reads your marks. It does not check a signature, test your site, or predict what any agent will do.

They grade different traffic. The AI Bot & Scraper Traffic Control Readiness Gate ($49) grades read traffic from crawlers and scrapers. Visiting Agent Rulebook grades agents that write and transact: book, buy, change and cancel, plus browse and quote. If your question is who may fetch your pages, start with the gate. If it is what a visiting agent may do once it arrives, start here.

RULEBOOK HOLDS AS DESCRIBED means no row is more open today than its layers support, no customer-rights review is open and your rules are published. “Holds as described” means the entered marks satisfy this product’s stated rulebook conditions. It is not a certification that the rules are lawful, complete, secure or technically enforced. RULEBOOK HAS GAPS means at least one row is treated more openly than its layers support, a refused change or cancel row has its customer-rights review open, or your rules are stated nowhere an outsider can read. UNVERIFIED MONEY PATH applies however the other rows read. A money-moving action is marked admitted while the identity layer is below this instrument’s verified mark. This result reflects only the user’s marks. RedHub did not inspect the agent, verify a signature, authorize a payment or determine whether the action was permitted. This self-assessment applies the stated rules to the user’s marks. RedHub did not visit the site, inspect traffic, verify a signature, identify an agent, validate user approval, interpret a contract, authorize a payment or confirm an outcome. The result is not legal, payment or security advice and does not determine whether any action is permitted or required.

REFUSE when Scope is 0, or when the action moves money and Identity is 0. ADMIT when Scope and Refusal are 2 and, for a money-moving action, Identity, Intent and Limit are 2 as well. REVIEW is anything else. If the action is change or cancel and today is refuse, the row displays REVIEW with a customer-rights review, whatever its layers read. Book, buy, change and cancel count as money-moving. Browse and quote do not.

It fires when you admit a money-moving action today and Identity is below 2 (verified). The roll-up then reads UNVERIFIED MONEY PATH however the other rows read. A money-moving action is marked admitted while the identity layer is below this instrument’s verified mark. This result reflects only the user’s marks. RedHub did not inspect the agent, verify a signature, authorize a payment or determine whether the action was permitted. It does not stop any transaction.

Only where a row is short. Gate rows come first, then gap rows, then open customer-rights reviews, then publishing your rules if none are stated. It names a row and the matching section of the Layer Guide in the download. At RULEBOOK HOLDS AS DESCRIBED it names nothing, because no row is more open than its layers support and no customer-rights review is open.

No. A successfully verified signature provides technical evidence that the request was signed using the private key corresponding to the published key used in verification. End-user approval, site permission, payment authority and outcome verification are separate questions. The association between a key, URL and operator depends on the applicable registration, directory and verification process. That is why the rulebook keeps Identity and Intent as separate layers: Identity asks what you checked about the agent's key or credential, and Intent asks whether you have a record of what the person approved. Web Bot Auth, an IETF working-group draft as of October 31, 2026, concerns signing agent requests. The rulebook does not check a signature and does not treat that draft as ratified. Not legal, payment or security advice.

Yes, as a mark. The published-rules mark runs 0 (stated nowhere an outsider can read), 1 (in your terms or a policy page) or 2 (also in a machine-readable file). Candidates include siteai.json from the A2WF W3C Community Group and a /.well-known/agent-policy.json community draft. Both are drafts as of October 31, 2026, and neither is ratified. The rulebook grades whether you have published, not which format you chose. Check a draft's current status before you adopt it. Not legal, payment or security advice.

No. REFUSE describes the site’s current treatment of that automated request. It does not determine whether refusing the channel is lawful, whether the customer has an applicable change, cancellation, refund or accessibility right, or whether another channel is sufficient. A refused change or cancel action triggers customer-rights review in this instrument. A refusal mark describes the site’s current treatment of the automated request; it does not determine that refusing the request is lawful or appropriate. For an action that changes or cancels an existing customer order, booking, account, subscription or service, record whether the customer retains an effective permitted way to exercise any applicable change, cancellation, refund or accessibility right. Applicable requirements vary by transaction, contract, jurisdiction and customer relationship. This instrument does not determine what rights apply or whether another channel is sufficient. Obtain qualified advice before relying on refusal as the completed policy for a customer-rights action.

No. It reads marks you enter about your own rules. It promises no behavior by any agent or platform, it does not test your site, and it is not legal, payment or security advice. Where a row reads short, you know which layer to work on first.

A workbook (Visiting-Agent-Rulebook.xlsx), a runnable Python engine with its config and a worked sample, the Layer Guide and Reading-Your-Rulebook playbooks as .docx files, and a README. It is $79 once, with lifetime access and 12 months of updates. Refunds are 30 days. Questions go to support@redhub.ai.