Redaction and Audit Logs for Agent Systems

This build lab focuses on the logging boundary that protects secrets while preserving accountability. The goal is not to copy a private machine setup. The goal is to learn the architecture pattern well enough to build a small, classroom-safe version.

Redaction should happen before logs leave the process, and audit logs should record action, actor, approval, result, and risk without exposing sensitive values.

Build the small version

1. 1Draw or write a redaction checklist and audit-log schema for agent actions.
2. 2Mark which parts are user-facing, which parts are internal, and which parts require approval.
3. 3Choose one low-risk workflow and implement only that workflow first.
4. 4Add one failure case before adding a second feature.
5. 5Write a short operator note: what the agent may do, what it must ask about, and what it must never do.

audit_event:
  time: 2026-04-27T12:00:00Z
  actor: student-demo
  action: send_email_draft
  approval: teacher_confirmed
  inputs_redacted: true
  result: draft_created
  risk: medium

Never log:
  - tokens
  - auth headers
  - private message bodies
  - service-role keys

The big idea: redaction is not decoration. It is part of the product architecture students need before an agent becomes safe enough to use with real people.