Refine & review
Per-section AI refinement and the four-agent review panel.
Two AI tools sharpen the spec: refine improves a single section on demand, and review sends the whole document to a panel of four expert agents that flag what's missing or weak.
Refine a section
Each section can be improved independently:
- Click Ask Modus to refine on a section.
- The composer switches to refine mode for that section.
- Pick a suggestion chip or type a free-text instruction.
- Modus rewrites the section and logs the change as a card in the chat thread.
Suggestion chips are per-section presets like "Sharpen scope boundaries" or "Add an edge-case persona," shown with a sparkle icon. A status chip animates through pending → updated (or error) and settles back to idle.
Every refine becomes a card in the chat thread recording the section, your instruction, the chip used, and the result, giving you a history of what changed and why.
Multi-agent review
Hit Review in the subheader to have four expert personas critique the spec in parallel:
| Persona | Lens |
|---|---|
| PM | Product strategy, prioritization |
| Designer | UX / UI, user experience |
| F&B Engineer | Frontend / backend feasibility |
| Web3 Architect | Contracts, on-chain architecture |
Findings land in the Review tab, each tagged with a severity (warning or info) and the section it's about. When more than one persona flags the same thing, you'll see co-attribution ("Designer agrees").
Applying findings
Findings can be applied in a batch: each one routes into its section's refine flow and runs in sequence, with status shown as it goes, so a review becomes a to-do list you can action in one pass.