Security & Data

Verification data should remain traceable, scoped, and reversible

This ledger separates what the workflow can record from what depends on your runtime and connected provider. It makes no claim of certification and treats privacy controls as operating decisions that must be verified before production use.

Data handling ledger

ControlExpected recordStatusReview gate
ProvenanceProvider, check type, result state, and observed timestamp remain attached to the candidate.Config-dependentReject exports that erase source or checked time.
Refresh cadenceEach result carries a freshness window selected for the campaign’s risk and timing.Operator definedRecheck stale records before reuse.
API credentialsSecrets stay in an environment store, outside prompts, repositories, exports, and screenshots.Inspect locallyUse minimum scope and rotate exposed keys.
Professional dataInputs are limited to relevant business identity and professional contact context.Purpose limitedDo not use personal or sensitive data for convenience.
RetentionTemporary candidates, logs, exports, and rejected results have named owners and deletion periods.Operator definedTest correction, suppression, and deletion paths.
Human reviewValid, catch-all, and unknown states remain visible before CRM or sending handoff.RequiredA reviewer approves relevance and lawful use.

Verification state is evidence about address risk, not proof of identity, consent, deliverability, or commercial interest. Catch-all domains may accept SMTP traffic without confirming an individual mailbox. Unknown results may reflect provider policy, throttling, greylisting, or transient network conditions. Preserve those distinctions so downstream automation does not convert uncertainty into confidence.

Operational trust checklist

  • Review the package identity, requested permissions, network destinations, and update path before execution.
  • Keep provider keys in scoped runtime secrets and prevent them from entering prompts or logs.
  • Document result definitions and the conditions that produce valid, invalid, catch-all, or unknown.
  • Apply suppression and opt-out records before a fresh technical verification result.
  • Test who can access, correct, export, retain, and delete professional contact data.
  • Separate mailbox verification from copy approval, authentication checks, and sending.

Inspect the controls in your own runtime

Begin with masked test records and a non-production credential. Verify data flow, errors, retries, retention, and deletion before connecting a real campaign.

npx -y @okki-global/okki-go-taroball