Single-Identity Business-Network Actor-Scope Automation: The Guard That Assumes an Organization Doesn't Work Here

Reusing a proven authorization pattern across every system in a codebase is usually the safer default — one well-understood mechanism is easier to audit than a dozen bespoke ones. That default stops being safe the moment a new system's underlying design genuinely doesn't match the assumption the pattern depends on. A guard built to resolve "who is asking" from an active user session works precisely as long as every caller has one; the first caller that doesn't — a background job, a service-role integration, an AI agent with no login of its own — doesn't fail loudly. It fails by resolving to nobody, unless something was deliberately built to handle that case differently.

The migration's own header comment states the distinguishing fact directly: Vértice is a single network with no organization boundary — identity is the member's own user account — and guards without an explicit argument rely on the currently authenticated session, which a service-role-driven caller like an AI agent simply does not have. Building this tool suite on the standard no-argument guard pattern used elsewhere in the codebase would not have failed with an error; it would have silently resolved to no valid actor, and every call would have failed in a way that looked identical to a real authorization rejection, making the two indistinguishable without reading the guard's own source.

The tool suite avoids that entirely by never relying on session resolution: every function takes the calling agent's own identifier as an explicit argument, looks up that agent's record directly, and — critically — resolves the human identity it should act as from a single, fixed value hard-coded into the context-resolution function itself, rather than from anything the agent's own request supplies. That fixed value represents the network's own house account: the one membership that represents the business itself inside this single shared network. Every tool built on top of that context — updating the house's own profile, registering its company, drafting or publishing an opportunity, submitting a referral, reading its own connections — inherits that same fixed identity automatically; there is no argument anywhere in the suite through which a caller could substitute a different member's identity instead.

Two further protections sit inside that same design. First, two specific profile fields — legal name and phone number — are excluded entirely from the agent's own profile-update tool; both are protected at the database layer by a trigger that the tool's own single write path (a shared, pre-existing save function used for every other field) does not bypass, meaning even a request that explicitly asked to change them has no code path capable of doing so. Second, because this network is new and had, at the time of writing, a single member profile in it, the training material built alongside these tools instructs the agent explicitly that near-zero counts across the panorama tool are a sign of early adoption, not a malfunction — a specific, concrete guardrail against an AI agent seeing an empty-looking network and reporting it to the business owner as though something were broken.

Applying the standard no-argument guard used elsewhere in the codebase to this system would mean every AI-agent call attempts to read the current session's own authenticated user — a value that simply doesn't exist for a service-role process. The call doesn't error out with a clear "no session found" message; it resolves to no valid actor, and every downstream check built on top of that fails the same way a genuine unauthorized request would, with nothing in the response distinguishing the two.

An AI agent can now operate on Centriu Vértice's single, shared business network on the business's own behalf specifically — never impersonating another member — using an authorization pattern built for a system that genuinely has no organization boundary, rather than one borrowed from systems that do.

This actor-scope pattern is specific to Vértice's own no-organization design — it does not apply to, and does not change, the organization-scoped authorization guard used by every other Centriu system, which continues to resolve identity from an organization boundary that genuinely exists for those systems.

Applying one proven authorization mechanism uniformly across every system in a codebase is easier to reason about — until one system's underlying identity model genuinely doesn't match what that mechanism assumes. Centriu's approach here is to recognize that mismatch explicitly rather than force a session-based guard onto a system built with no session-based identity to check, building an actor-scoped alternative specific to what this one system actually is.