Build and run a quality gate skill using custom AI skills
Marketing teams rely on rules and governance processes to ensure audiences are activated correctly. Before launching an audience to a destination, teams often need to verify suppression lists, frequency caps, consent requirements, and naming conventions.
The challenge is that these checks frequently depend on tribal knowledge and manual reviews. When processes live in people’s heads, mistakes can happen.
In this video, you’ll see how a custom Coworker skill acts as an Activation Gate, automatically validating audiences against your organization’s activation standards before they move downstream.
Sample activation quality gate skill
You can create your own reusable Activation Quality Gate skill by pasting a prompt into Coworker. Coworker’s skill-authoring capabilities convert the prompt into a saved skill within your own environment. A sample based on the video demonstration follows.
The key is to define your own pass/fail standards for the three governance gates:
- Suppression / Consent
- Frequency Cap
- Naming Convention
The framework remains the same for everyone. Customize the sections marked with[...]to match your organization’s standards.
Master prompt
Save this as a skill called “Activation Quality Gate.”
It's a governance gate that runs a pre-activation checklist before any audience is sent to a destination.
It is read-only. It never activates, mutates, or copies anything.
Resolve the named audience and destination from our Knowledge Graph, evaluate the three gates below, then render one visual scorecard containing:
- An Alert banner
- One MetricCard per gate
- A DataTable with:
- Gate
- Status
- Finding
- Required Fix
Provide a single verdict:
- CLEARED only if all three gates pass
- BLOCKED if any gate fails
For every failed gate, provide the specific remediation needed.
All gates fail closed:
- Missing data = BLOCKED
- Never assume success when information is unavailable
Trigger phrases:
- "run the activation gate"
- "is this audience ready to activate"
- "pre-activation checklist"
- "can I activate to ..."
The three gates are:
[Paste Gate 1, Gate 2, and Gate 3 definitions here]
Gate 1: Suppression / Consent
Edit this section to match your organization’s suppression and consent requirements.
Gate 1 – Suppression List
Pass only if a recognized suppression, opt-out, or consent audience is applied alongside the target audience.
Discover eligible lists using name patterns such as:
- suppress
- opt-in
- opt out
- consent
- do not contact
Because suppression lists may live in destination dataflows rather than audience metadata, require the marketer to confirm one is attached.
If no suppression or consent list exists anywhere in the sandbox, fail hard.
Our standard:
[Example: A consent audience is mandatory for all email and SMS destinations. For direct mail destinations it is optional.]
Gate 2: Frequency Cap
Edit this section to match your organization’s delivery-frequency requirements.
Gate 2 – Frequency Cap
Read the delivery frequency on the resolved destination.
Pass if:
- Frequency is present
- Frequency is bounded
Fail if:
- Frequency is blank
- Frequency is unbounded
Our standard:
[Example: Frequency must be DAILY or less frequent. Any hourly cadence or blank value is blocked.]
Gate 3: Naming Convention
Edit this section to match your organization’s audience naming rules.
Gate 3 – Naming Convention
Evaluate the audience name programmatically.
Any rule violation causes failure.
Block names that:
- Contain "test"
- Contain "copy"
- Contain an auto-copy suffix such as _[6-hex]
- Contain timestamps
- Contain 24-character object IDs
- Start with a bare number or cryptic short code
- Are entirely lowercase
- Are excessively short or unclear
- Use generic defaults such as:
- Save audience
- Email
- New Accounts
- Lack a category–qualifier separator
Our standard:
[Example: [Line of Business] – [Criteria] in title case]
Example:
Mortgage – High Propensity Prospects
When blocked on naming, always propose a compliant replacement name.
Guidance
1. Customize only the bracketed sections
Only update only the sections contained in [...].
These sections define the organization’s specific governance standards.
Everything else should remain unchanged:
- Audience resolution
- Gate evaluation
- Scorecard rendering
- Verdict logic
2. Verify Prerequisites
This skill depends on:
- Knowledge Graph access
- Audience discovery
- Destination discovery
- Suppression-list discovery
- Visual artifact support
- Alert banner
- MetricCards
- DataTable rendering
If these capabilities are not available in the customer’s environment, the skill cannot run as designed.
3. Keep the Skill Read-Only
The skill should always remain read-only.
Include this requirement explicitly in the prompt to ensure the skill is never confused with an activation workflow.
The Activation Quality Gate evaluates activation readiness only. It does not activate audiences, modify configurations, or copy data.