Integrity
Integrity

Agent-assisted remediation

Use assistant previews for Integrity findings while preserving redaction, stable codes, and human approval.

Integrity can help you explain and organize findings before you fix them. The assistant preview on the Findings tab consumes the same display-safe diagnostics you see in the table.

It is a planning aid, not an apply button.

Explain selected findings Admin Editor

  1. Open Workspace -> Integrity -> Findings.
  2. Select one or more findings.
  3. Click Explain.

The explanation groups selected findings by rule pack and severity, includes stable finding codes, and suggests the next remediation targets.

Draft a proposal preview Admin Editor

  1. Select one or more findings.
  2. Click Draft proposal.
  3. Review the proposal preview.
  4. Follow the linked remediation surface if you decide to fix the issue.

The preview marks the draft as Review required and No evidence mutation. It links to existing remediation targets, but it does not apply changes.

What agents may do

Agents may:

  • explain why selected findings matter
  • group related findings
  • identify likely root causes from display-safe evidence
  • draft a human-reviewable remediation plan
  • link you to existing remediation surfaces

Agents must not:

  • silently mutate requirements or evidence
  • create missing trace links without review
  • rewrite identifiers
  • change imports or baselines
  • approve reviews
  • bypass source restrictions
  • reveal restricted detail in summaries or proposals

Restricted findings are rewritten before assistant input is built. Hidden source names, artefact identifiers, link details, import evidence, coverage config context, and secondary actions are removed.

Restricted findings

If a selected finding includes restricted detail that you cannot access, the assistant preview shows a redaction notice. The finding is represented as a generic restricted finding with a route back to the redacted Integrity filter.

To perform a detailed repair, a user with the required restricted source or entity access must review the underlying issue.

Future richer agents

Future agent runtime work may stream richer explanations or create durable proposal records. Those flows must keep the same contract:

  • consume display-safe Integrity diagnostics
  • preserve stable finding codes
  • preserve remediation modes
  • keep restricted details redacted
  • require explicit human approval before any source-of-truth mutation

On this page