Auditor Code of Conduct

An auditor must not sign an audit where they have a conflict of interest, including but not limited to:

• You authored or co-authored the listing under review.
• You are employed by, contracted to, or financially dependent on the listing’s creator or their affiliated entities.
• You hold a material economic interest (greater than 1% of issued capital, or any director / officer position) in the creator’s business.
• A close family member is the listing’s creator.
• You have received any consideration from the creator (cash, gifts, services, equity, future commitments) outside the standard auditor payout flow operated by Trust Agent.

If a conflict arises mid-audit, decline the claim and notify info@trust-agent.ai. The audit will be re-assigned.

Before signing an audit you must, at minimum:

• Read the listing’s prompt material, behaviour spec, and capabilities manifest in full.
• Review the automated security scan output, including all FAIL and WARNING findings, and form a view on whether each is correctly classified.
• Check that the listing’s claims (description, category, declared limits) match the observed behaviour.
• Document any findings the automated pipeline missed in the narrative.
• Set the human-review score honestly, on the rubric published in the auditor handbook (currently: 90+ exceeds bar / 80-89 meets bar / 70-79 marginal / below 70 fails).

Sign-off attests to your honest professional judgment that the listing is fit to carry the resulting badge.

You may see proprietary creator material during an audit (full prompts, internal documentation, capability manifests). Treat it as confidential:

• Do not republish, share, or train models on the material.
• Do not use the material to compete with the creator outside the audit.
• Do not disclose to third parties without the creator’s written consent or a lawful order.

The confidentiality duty survives termination of your auditor status.

An auditor must not:

• Sign off on a listing they have not personally reviewed (no “rubber-stamp” claims).
• Inflate or deflate the human-review score in exchange for any consideration.
• Suppress findings the creator finds embarrassing.
• Coordinate with other auditors to bias outcomes for or against any creator.
• Use AI tools to generate the audit narrative in a way that obscures the absence of human review.

The auditor payout (currently 50% of the creator’s audit fee, subject to change in the platform configuration) is paid into your declared Stripe account within the payout cycle published in the auditor handbook. Payouts may be withheld pending investigation of any AUP or Code breach.

Payouts are gross of tax. You are responsible for declaring auditor income to the relevant tax authority.

Every claim, narrative edit, score change, sign-off, and rejection is logged with your auditor ID and timestamped. The log is part of the public audit certificate at /audits/{id}. Trust Agent reserves the right to re-audit any listing you signed at any time and, where the re-audit materially diverges from your sign-off, to investigate.

Auditor status may be suspended (with stop on payouts) pending investigation, and removed permanently for material breach. Removal is recorded on the auditor’s profile and may be referenced on listings they previously audited where their sign-off is the operative human review.

If you suspect another auditor has breached this Code, or that platform staff have pressured you to compromise an audit, contact info@trust-agent.ai. We treat whistleblowing reports confidentially and will not retaliate against an auditor for a good-faith report.

We may update this Code as the audit programme matures. Material changes are communicated to active auditors at least 30 days before they take effect. Continued audit activity after the effective date constitutes acceptance.

AgentCore LTD, 20 Wenlock Road, London, England, N1 7GU. Email: info@trust-agent.ai.