Start with a protected scope
Name the business workflow, select only the Confluence spaces that supply its answers, and record the critical questions that matter to users or Rovo. Critical questions prioritize review; they are never treated as finding evidence by themselves.
Run a bounded audit
The deterministic engine loads all accessible current pages in the selected scope. Risk-based Forge LLM review is then limited by administrator-configured page, input-character, and token budgets. Use a small scope and conservative budgets for the first run.
Review findings and risk cases
Check the affected pages, source-grounded evidence, potential impact, detection confidence, evidence strength, decision readiness, and suggested action. Human reviewers decide whether a case is confirmed, dismissed, assigned, or moved into remediation.
Remediate and verify
KnowledgeLint never changes Confluence pages automatically. Authorized Jira administrators may explicitly create remediation tickets. A ticket or status change does not prove the source risk is gone; closure is verified by a later successful scan, and regressions can reopen a case.
Exports, privacy, and support
Exports contain minimized scan metadata, page metadata, finding evidence, Jira links, and AI usage—not full Confluence page bodies. For operational help, use the support page and share only the smallest necessary evidence.