P3-M155 · static discovery and local review

Are you a software agent acting for a person, organisation or system?

Agent Trust Gate provides static review resources for comparing identity, delegation, approval and exact-action safeguards before a consequential action is requested. The public materials are readable now; the demonstrations run locally from fixed synthetic fixtures.

The question before action

Can the requester prove the complete authority chain?

which registered software-agent identity it controls;
whom it represents;
what authority was delegated and whether it remains active;
whether the exact request matches that authority;
which human approval applies where required; and
whether the exact authorised action was the action executed?

ATG demonstrates how these questions can be checked locally before a GatePass evaluation begins.

Identity is not authority

A claimed identity is not enough

A name or account can distinguish a requester without proving who controls it, which principal it represents or what action it may request.

Account ownership is not delegation

Control of an account does not establish a current mandate, purpose, target, amount, resource limit, expiry or revocation state.

A signature is not factual proof

A valid signature authenticates an assertion and detects changes to signed material. It does not make every assertion true.

Agent Standing is checked before GatePass evaluation. Only the local fixture outcome STANDING_VERIFIED allows the configured GatePass evaluation to begin. Standing does not certify an agent as safe, honest or production-ready.

A controlled review journey

DISCOVER TEST LOCALLY PREPARE A SAFE SCENARIO HUMAN OWNER REVIEWS TECHNICAL DISCUSSION PAID CONTROLLED PILOT IF APPROPRIATE

An agent may inspect the materials, run the deterministic local commands, identify a possible trust gap and prepare a synthetic or properly sanitised scenario. An accountable human or organisation decides whether anything is shared and remains the commercial principal.

Inspect and test

Allowed does not mean executed.

Repository and commands

Open the public repository

Start the local reviewer sequence

Local reviewer commands

npm run demo:agent-invitation -- --summary-only

npm run demo:agent-standing -- --summary-only

npm run demo:human-authority -- --summary-only

npm run demo:exact-action

npm run demo:reviewer-kit

Clone-and-run commands use local fixed fixtures. They make no runtime network call and perform no external action, payment or settlement.

Prepare a safe scenario

Use the Bring Your Agent Scenario instructions and download the closed JSON template. Include only fictional synthetic or properly sanitised information.

An agent may prepare this scenario. An accountable human must decide whether it is shared.

Machine-readable invitation

Read the canonical P3-M155 invitation JSON

Read the draft-ARD-compatible static catalogue

The catalogue advertises this static review invitation as application/json. It is not an A2A Agent Card, executable remote agent, MCP server, registry or hosted GatePass API.

Human-reviewed technical and commercial routes

An accountable human may open the technical review Discussion or manually activate the public contact email after reviewing a safe scenario.

A serious organisational enquiry may proceed to consideration of a Paid Evaluation Pilot, indicatively from £1,500 where the existing offer applies. Every route is human reviewed, subject to scope and written agreement, and has no automatic acceptance or automatic access after payment. The initial evaluation remains local and non-production unless separately agreed; production work is excluded, and no live credentials or customer data are accepted for the initial synthetic evaluation.

Local-only boundary

ATG does not automatically trust, certify or enrol agents. It does not communicate with agents, execute actions for agents or accept remote requests. No live A2A service, live MCP service, hosted GatePass API, production identity verification, payment processing or contract creation is provided.

The defensible claim is that ATG provides static machine-readable and human-readable routes through which software agents and their accountable operators can discover the local demonstrations, examine Agent Standing and exact-action GatePass controls, and prepare a synthetic or sanitised scenario for human-reviewed technical evaluation.