AI Knowledge Hub
Menu

Adtech Agents / Experimentation

ChatGPT Ads Experiment Supervisor

Coordinate task-led ChatGPT advertising experiments across creative, policy, measurement ownership, platform inspection, reconciliation, and memory.

Setup requiredAgent workflowRequirements & limits

Inputs

The released example shows these fields. Check the guide for which are required.

  • mode
  • brand
  • objective
  • task
  • risk_policy

Outputs

  • Experiment plan
  • Readiness decision
  • Reconciled evidence
  • Experiment memory record

Before you begin

  • Install the agent and begin with its mock workflow; use the compiled contract to configure connected providers.

Try this first

Review the coordinated workflow and its runtime requirements before configuring an agent app.

Limits for this step
  • Provider authentication and the pinned control-plane dependencies must pass readiness before connected use.
  • The supervisor is advisory and does not execute campaign writes.
  • May plan, inspect approved read-only data, and reconcile supplied evidence; live campaign writes require the existing policy gate, executor, and human approval.
View agent workflowHide agent workflowv0.2.0

Experiment supervisor for task-led conversational advertising and agent-mediated commerce.

What it coordinates

  • Task hypothesis design
  • Creative and compliance readiness
  • Signal ownership and measurement
  • Mock or read-only platform inspection
  • Event reconciliation and experiment memory

Autonomy and approvals

Advisory

May plan, inspect approved read-only data, and reconcile supplied evidence; live campaign writes require the existing policy gate, executor, and human approval.

Read agent instructions (AGENT.md)

ChatGPT Ads Experiment Supervisor

Identity

Advisory supervisor for task-led ChatGPT advertising and agent-mediated commerce experiments.

Mission

Coordinate task design, creative evaluation, signal ownership, policy controls, read-only platform inspection, independent outcome reconciliation, and experiment memory.

Modes

  • plan-only: create the experiment and governance plan.
  • mock-run: use local adapter and event fixtures.
  • read-only: inspect approved account data without mutations.
  • approval-gated: prepare bounded change plans for the existing executor.

Workflow

  1. Use marketing/task-map-designer to define task hypotheses.
  2. Route creative work through marketing/creative-operating-system-supervisor and compliance evaluation.
  3. Use adtech/advertising-signal-ownership-auditor before activation.
  4. Inspect mock or live read-only data through adtech/openai-ads-api-client.
  5. Use the existing policy gate and executor for any proposed write.
  6. Reconcile Pixel, server, CRM, order, and reversal events with adtech/conversion-event-reconciler.
  7. Use marketing/agent-share-of-choice-evaluator to evaluate paid, earned, and executable routes.
  8. Write a versioned experiment record conforming to shared/schemas/advertising/experiment-record.schema.json.

Output

  • task and hypothesis
  • creative and policy readiness
  • signal ownership ledger
  • activation or mock-run plan
  • reconciled evidence
  • share-of-choice evaluation
  • experiment memory record
  • next decision

Guardrails

  • Never hold platform credentials.
  • Never execute writes directly.
  • Do not activate an experiment without a measurement ledger.
  • Keep paid placement separate from earned recommendation.
  • Treat platform conversions as one evidence source, not final truth.
  • Require human approval for activation, budget, targeting, feed, and tracking changes.

Summary sources · Catalogue metadata, agent.yaml, AGENT.md, examples/input.json, examples/output.json · v0.2.0 · Release package

Install agent files

These commands install package files and generate runtime configuration where supported. Confirm discovery in your agent app, then check configuration, authentication and a first useful operation.

Install (Codex)

./bin/skills-hub install --module agents --entry adtech/[email protected] --runtime codex

Install (Claude)

./bin/skills-hub install --module agents --entry adtech/[email protected] --runtime claude

Install (Generic)

./bin/skills-hub install --module agents --entry adtech/[email protected] --runtime generic --target ./my-agent/agents

How You Can Use This

Setup required

Resource form: Agent workflow

Configuration, credentials or runtime bindings are required. Check the prerequisites and limits for your intended environment.

First run: Install the agent and begin with its mock workflow; use the compiled contract to configure connected providers.

Current limits

  • Provider authentication and the pinned control-plane dependencies must pass readiness before connected use.
  • The supervisor is advisory and does not execute campaign writes.
What installation and verification mean
File installation
Places package files and generated configuration in the selected directories.
Native discovery
Your agent app finds and accepts the instructions, tools or plugin registrations.
Mock execution
Exercises sample fixtures. Connected data and provider effects need a separate check.
Connected operation
Requires current configuration, provider authentication, scoped permissions and a successful operation in your intended environment.

Declared compatibility

Compatibility unknown: versions and capabilities have not been declared.

Classification: declared

Technical details and release files

Review and lifecycle

Review maturity: Experimental

Security review declaration: Not declared

Lifecycle: Active

These release declarations are separate from evidence that a workflow operates in your environment.

Runtime & Dependencies

ID: adtech/chatgpt-ads-experiment-supervisor

Tags: chatgpt-ads, experiment-supervisor, measurement, agentic-commerce, governance

Runtimes: codex, claude, generic

Skill dependencies: 5

Agent dependencies: 2

Tool dependencies: 3

Dependencies

Skills: marketing/task-map-designer, marketing/ai-output-eval-scorecard, adtech/advertising-signal-ownership-auditor, marketing/agent-share-of-choice-evaluator, agentops/ad-platform-policy-gate-designer

Tools: adtech/conversion-event-reconciler, adtech/openai-ads-api-client, adtech/ad-platform-executor-template