AgentScape
← Directory
SE

Composite agents

Secretary Policy Supervisor

AgentScape Patterns
AgentPatternListedNot monitored

Supervisor agent for a secretary team that reviews proposed actions against approval rules, tool grants, risk level, and escalation policy.

Pattern catalog entry

First-party AgentScape reference pattern. It is a reusable definition, not an independent vendor listing. The catalog-origin label is separate from the Listed, Verified, or Trusted badge, which only reports recorded technical evidence.

A2ARESTassistivemedium riskmicro budget

Before you install

Inputs, access, schedule, and visibility

Review what this agent needs and what it can access. Connected services are not granted automatically; the person running the agent must connect and authorize their own account.

What you provide
Child-agent proposal, requested tool, data class, action level, user approval rules, and recent context summary.
What you receive
Decision JSON with allow, deny, or escalate, reason, required approval, and safe alternative.Output mode: plain text
Execution
This release has no executable connector for the external requirements below. The agent can only use data you paste into the task; it cannot fetch or update those systems.
Cadence and timezone
Manual, on-demand runs. No daily frequency was declared.

Runtime tool access

No runtime tools are declared.

Additional publisher-declared requirements

  • policy registry
  • audit log

Required connected services

No supported AgentScape connector maps to the external requirements listed at left. AgentScape will not access or update those systems.

Not available in this AgentScape release: policy registry, audit log. The agent cannot read from or write to those systems. Paste only the relevant data into the task, or choose an agent whose required integration is supported.

Personal data and workspace visibility

This agent may process general, personal, customer. Run inputs and saved outputs can be visible to workspace members who can manage agents or runtime settings. Review workspace membership before using a personal account.

micro budget tier. The cadence is a usage estimate, not a price quote. Actual provider cost depends on the selected runtime and model, input size, retries, failures, and workspace budget policy.
Directory listing onlyReview workspace setup

This listing does not have an immutable runnable definition. It remains visible for discovery but cannot be installed.

Try it — chat with Secretary Policy Supervisor

Sandboxed demo
Demo boundary: this chat cannot read your Gmail, Calendar, Drive, or local files and cannot take external actions. Use fictional sample data here; install the agent and connect the required services for a real workspace run.

Start a conversation to see how Secretary Policy Supervisor thinks and decides.

Replies are either model-generated from this demo conversation or explicitly marked fictional illustrations. Neither is evidence that a connected tool ran or that an external action succeeded.

Trust & verification

Methods disclosed
Listed

Submitted and passed automated policy review. No technical verification has been performed yet.

  • Identity verified

    Publisher controls the declared domain.

  • Endpoint reachable

    Agent endpoint responds over public HTTPS.

  • Signed manifest

    No manifest signature declared.

Health

Endpoint monitoring
30-day uptime

This agent has no public endpoint to monitor.

Purpose & governance

Approve, deny, or escalate child-agent proposals before any sensitive or write-level action proceeds.

Success metric: Every risky proposal receives a clear allow, deny, or escalate decision with an audit reason.

Human-in-the-loop
Yes
Observability
Yes

Capabilities & policy

Capabilities

  • policy review
  • approval routing
  • risk classification
  • audit summaries

Data classes

  • general
  • personal
  • customer

Allowed uses

  • policy review
  • approval routing
  • risk classification
  • agent handoff review

Prohibited uses

  • rubber-stamp approval
  • bypassing grants
  • hiding denied actions
  • approving destructive actions without authority

Auth methods

  • OAuth
  • Service Account

Deployment

  • Cloud
  • Private cloud

Links & resources

Lifecycle

Review, verification, and enforcement history
  1. APPROVED

    Listing entered the public registry.

Reviews & ratings

No moderated reviews yet.

Share your experience