The same underlying Claude capability, described in two professional vocabularies. Select any capability to see both perspectives.
Capabilities that appear across multiple roles
Auditor
Programmatic Control Points That Cannot Be Bypassed
This is the enforceable control point that separates a policy statement from a verified control. The gate exists in code, not in instructions.
Full entry on Auditor page →Marketer
Pre-Publication Stops That Cannot Be Talked Around
FTC disclosure requirements, industry advertising restrictions, and platform policy violations do not wait for someone to remember to check. This gate exists in the architecture. It fires before the content moves.
Full entry on Marketer page →Auditor
Scope Containment — The System Does Only What Was Approved
This is your documented boundary. What is not in the allowlist cannot be executed. The AI cannot wander outside its defined environment regardless of what it is asked to do.
Full entry on Auditor page →Marketer
Approved Claims Only — No Freelancing
This is the control that keeps AI-generated content from inventing product capabilities, misrepresenting competitive positioning, or making claims Legal has not reviewed. The system cannot reach outside its approved sources regardless of what it is asked to produce.
Full entry on Marketer page →Auditor
Consistent, Reproducible Outputs Under Identical Conditions
If you cannot reproduce an output, you cannot verify it. Structured outputs enforce a fixed schema so that results are predictable, comparable, and defensible under scrutiny.
Full entry on Auditor page →Marketer
Brand Voice That Holds at Volume
This is the architectural answer to brand drift. The system does not decide what format to use — it follows a schema your team defined. What the schema does not permit, the system cannot produce.
Full entry on Marketer page →Auditor
Evidence Integrity — The System Cannot Audit Itself
Your evidence sources and your AI system are architecturally separate. The system being evaluated does not control the record of its own behavior.
Full entry on Auditor page →Marketer
Your Brand Assets Stay Yours
Your brand voice is intellectual property. The question of whether it can be used to train a model that serves your competitors is a legal question, not a technical one. ZDR is the contractual and architectural answer that closes that exposure.
Full entry on Marketer page →Auditor
AI Assistance Disclosure Built Into the Methodology
Emerging audit standards require disclosure of AI involvement in findings and recommendations. This is not optional. The output record makes that disclosure automatic and consistent rather than dependent on human memory.
Full entry on Auditor page →Marketer
AI Content Disclosure Built Into the Output Record
Disclosure is becoming a legal requirement, not a best practice. The output record makes that disclosure automatic and auditable rather than dependent on human memory at campaign velocity.
Full entry on Marketer page →Auditor
Model Version Control as a Change Event
A model update that changes system behavior is a change event under any IT governance framework. It must be documented, tested, approved, and logged — and this architecture supports exactly that requirement.
Full entry on Auditor page →Marketer
A Content System That Knows When It Changed
A model update that changes how your system writes campaign copy is a change event. It should be documented, tested against brand standards, and approved before it touches production output. This architecture supports that requirement.
Full entry on Marketer page →Role-specific capabilities
Auditor
Separation of Duties Enforced in Architecture, Not Policy
The allowedTools parameter defines exactly which actions each agent or role can take. The system cannot be talked into exceeding those boundaries.
Full entry on Auditor page →Marketer
Not yet documented for this role.
Auditor
A Complete, Tamper-Evident Activity Record
Session IDs tie every action back to a single originating request. The log is the evidence. It does not need to be reconstructed after the fact.
Full entry on Auditor page →Marketer
Not yet documented for this role.
Auditor
Human Review Gates That Are Visible When Skipped
A control that was skipped sixty percent of the time is a finding. This architecture makes skipped reviews visible in the log — they cannot be hidden by the system or by user behavior.
Full entry on Auditor page →Marketer
Not yet documented for this role.
Auditor
Not yet documented for this role.
Marketer
Examples That Teach Brand Voice, Not Just Describe It
Abstract brand guidelines produce generic content. Concrete examples produce brand-consistent content. The more precise your examples, the less rewriting your team does.
Full entry on Marketer page →Auditor
Not yet documented for this role.
Marketer
Brand Voice Packaged Once, Deployed Everywhere
A brand voice that has to be re-described for every new campaign is a brand voice that will drift. A skill is the packaging that makes your brand voice a controlled asset rather than a set of instructions someone has to remember to include.
Full entry on Marketer page →Auditor
Not yet documented for this role.
Marketer
Personalization That Knows Where the Privacy Line Is
GDPR, CCPA, and personalization at campaign scale are a collision waiting to happen without architectural controls. This is the control that keeps the personalization engine from reaching data it was not authorized to touch.
Full entry on Marketer page →Auditor
Not yet documented for this role.
Marketer
Cost Visibility Before Campaign Scale Hits
A system that works beautifully in testing can become expensive in production when a global campaign runs through it at full volume. Cost visibility is a campaign planning requirement, not a developer concern.
Full entry on Marketer page →