Ad Platform Policy Gate Designer

Ad Platform Policy Gate Designer

When to use

Use this skill when a team needs concrete governance rules for an agent that can propose or execute ad-platform changes.

Typical triggers:

• "What Google Ads actions can this agent auto-execute?"
• "Which DV360 changes need human approval?"
• "Design policy gates for bid, budget, and targeting changes."
• "Create deny rules for regulated categories or sensitive markets."
• "Turn our campaign-risk policy into executable thresholds."

Inputs required

Collect:

• Platforms and entity types: account, advertiser, campaign, insertion order, line item, ad group, keyword, audience, feed item.
• Action types: bid, budget, status, targeting, audience, creative, feed, tracking.
• Business context: client, market, spend tier, regulated category, SLA.
• Existing approval owners and escalation paths.
• Risk thresholds: budget delta, bid delta, audience expansion, geo expansion, pacing variance.
• Rollback and audit requirements.

Workflow

1. Classify actions by risk.

   • Low: read-only, recommendation-only, status summaries, small reversible bid changes.
   • Medium: bounded bid/budget edits, narrow targeting edits, non-publishing feed drafts.
   • High: budget increases, targeting broadening, audience expansion, live publishing, tracking changes.
   • Blocked: cross-client actions, unscoped exports, forbidden categories, policy overrides by agent.

2. Convert risk into gates.

   • Auto-execute only when action is low risk, scoped, reversible, and below thresholds.
   • Route medium/high risk to named approvers.
   • Deny actions that violate scope or organisational policy.

3. Define bounded execution grants.

   • Include workspace, platform, advertiser, entity, allowed action, max delta, expiry, and approval id.

4. Define audit evidence.

   • Require current state, proposed diff, policy decision, approval decision, execution payload, post-write state, and rollback patch.

5. Produce implementation-ready policy tables.

Output format

# Ad Platform Policy Gates

## Risk model
| Risk tier | Definition | Examples |
| --- | --- | --- |

## Policy gates
| Action | Auto-execute | Approval required | Deny when | Audit evidence |
| --- | --- | --- | --- | --- |

## Thresholds
| Metric | Low-risk range | Approval threshold | Block threshold |
| --- | --- | --- | --- |

## Bounded grant fields
- workspace_id
- platform
- advertiser_id
- entity_type
- entity_id
- allowed_action
- max_delta
- expires_at
- approval_id

## Engineering notes
1. ...

Guardrails

• Do not approve or execute live campaign changes.
• Do not make policy less strict than the user's stated organisational baseline.
• Treat missing client/workspace/advertiser scoping as a blocker for auto-execution.
• Require approval for targeting expansion, budget increases, tracking changes, and regulated-category changes by default.
• Recommend denial when rollback or post-write verification is unavailable for risky writes.

---
name: ad-platform-policy-gate-designer
description: Design approval gates, auto-execution thresholds, deny rules, and audit requirements for DV360, Google Ads, and ad-platform agent writes.
---