This mini-Canon is the short path for Eval Labs evaluators. It explains the evaluator workflow without requiring a first-time evaluator to read the full main Canon.
Before anything else
Eval Labs has passed the AI-reviewed platform readiness gate. That means the platform lifecycle has been proven: runs can be created, Lucia responses can be captured, reviews can be generated and persisted, and runs can be finalized. It does not mean Lucia has human quality approval. Human review remains the real Lucia behavioral-quality layer. Your job is to help judge whether Lucia is useful, trustworthy, clear, and calm for real human/operator use.Your first path
Evaluator access is broader than tester access, but it is still scoped. Use:- Custom Prompt Test
- Auto-generated Prompt Test
- Guest Facing Agent Verification Check
- Verification Results
- Controlled Batch Runner when assigned
- your own run, review, and history routes
- finalization for your own assigned runs
This section does not launch onboarding
This section is onboarding material. It does not mean Lucia is human-approved, that your account has been cleared for every Eval Labs surface, or that active workspace polish is finished. Your access and assignment still come from an owner or admin.Reading order
- What Eval Labs Is
- Your Role and Access
- AI-Reviewed vs Human Review
- Running Your First Custom Eval
- Reviewing Lucia
- Good Feedback Examples
- What Not To Do
- First Assignment Checklist
- Getting Help