Give an AI agent the permissions needed for its assigned job, enforce those limits in its tools and require evidence of the result before reporting completion. Separate diagnosis from actions that change systems, with explicit approval or escalation where needed. Helm demonstrates restricted roles and sandbox operations; its console source also exposes a completion-reporting gap.
We explored this in Helm, a demonstration built for Google's All Things Agentic Hackathon. It gives different agents different responsibilities over a product fleet, with disruptive drill actions confined to designated sandbox services.
Useful delegation gives an agent a job, a boundary and a way to prove completion.
WORKED EXAMPLE / HELM DRILL
Where does its authority stop?
Send one request. The boundary lives in the tools each role holds, not in how persuasive the request sounds.
- 01RouteAssigns the job
No tools - 02ReadWatch officer
Health checks · logs · read-only - 03ActEngineer
Ingress on/off · drill services only
target ∈ drill servicesallowedAn allowed action still needs checking.The engineer holds an ingress tool, and the tool accepts drill services only. The request is in scope. Report success only after the service answers again from outside; report a timeout otherwise.
- Request accepted
- Read the resulting state
- Report done
Explanation and primary sources
Three requests pass through three roles. One role routes the work and holds no tools. The watch officer reads evidence with read-only tools. The engineer can switch a service's public ingress, and that tool accepts drill services only. A request to bring the drill service back is allowed; the recommended check reports success only after the service answers again from outside, and a timeout otherwise. A request to take production offline is refused by the tool and escalated to a person, however it is worded. An error page instructing the agent to report the service healthy is read as evidence: its instruction text is quarantined and recorded, and the status code is kept. Recommended completion check, illustrated with Helm's roles; Helm is a contest build rather than a client case, and its own console route can report success after a timeout.
- Helm source
- Helm console completion check (pinned source)
- Helm recorded demonstration
- Your error page is a prompt
Make the division of work real
Helm has three roles: a commander routes the work, a watch officer reads the evidence, and an engineer acts through a restricted set of tools.
The distinction lives in the available operations. The diagnostic role has no remediation tools. The engineer's disruptive operations check which service is being targeted. A request outside that allowance cannot acquire permission just because the model finds it persuasive.
That is a useful question to ask of any proposed agent workflow: can we point to the place that enforces its authority?
You can watch the recorded Helm demonstration and inspect the source. It is a contest build and engineering example, not a commissioned client case study or proof of universal resilience.
Test the evidence path too
In the drill, a broken service returned an error page containing instructions to ignore the outage and report that it was healthy. That text arrived through a monitoring result, precisely where the agent was looking for evidence.
The demonstration screens recognisable instruction patterns, records the quarantine and preserves factual signals such as the response status. This is one defensive measure. A pattern filter cannot catch every prompt injection, so restricted tools remain essential even when the input looks ordinary.
The broader lesson travels beyond infrastructure. A ticket, document or incoming email can contain useful information and untrusted instructions in the same body. Reading it must not silently grant it authority over the workflow.
Completion needs a read-back
An API accepting a request does not prove that the operation finished. Helm illustrates the intended check, but its inspected console route has a gap: after twenty unsuccessful probes it can still report “live confirmed”. Completion reporting needs to depend on the observed result, including failure or timeout.
In another business, the equivalent could be reading a saved record back, checking whether a scheduled job finished, or confirming that a change affected the intended account. The exact check belongs to the task. The operator should be able to inspect it.
This is how we would scope a first engagement: one job, the data it may read, the changes it may make, its escalation conditions and the evidence required at the end. Broader autonomy follows demonstrated need and verified behaviour. WebMCP actions on the page bring that boundary into a user-facing product.
Before an agent touches real tools, ask to see one allowed action, one refused action and one failed action handled honestly. A convincing success path leaves those questions open.
The technical account has the detail for whoever will own the integration.
