Five expert personas independently analyze a proposed change and debate trade-offs before you write any code. Catches architectural, security, performance, and UX issues early.
Predict simulates a design review by running your proposal through five distinct perspectives: Architect, Security Expert, Performance Engineer, UX Specialist, and Devil’s Advocate. Each persona analyzes independently, then conflicts are identified and resolved into a clear verdict.
- Before implementing a major feature
- Before risky architectural changes (database migration, auth rewrite)
- When evaluating competing approaches
- Before touching production-critical code paths
| Persona | Focus |
|---|
| Architect | System design, coupling, extensibility, API contracts |
| Security | Attack surfaces, auth flows, data exposure, OWASP risks |
| Performance | Latency, memory, N+1 queries, caching, scalability |
| UX | User impact, cognitive load, accessibility, error states |
| Devil’s Advocate | Hidden assumptions, edge cases, what could go wrong |
| Argument | Description |
|---|
description | Feature or change proposal (natural language) |
--files <glob> | Optional glob to read affected code areas |
/ck:predict "Add WebSocket support for real-time notifications"
/ck:predict "Migrate authentication from JWT to session cookies"
/ck:predict "Replace REST API with GraphQL" --files src/api/**/*.ts
- Read — parse proposal and optional file paths
- Analyze — each persona reviews independently
- Agreements — identify points where 4+ personas align
- Conflicts — surface competing concerns between personas
- Resolve — weigh trade-offs for each conflict
- Verdict — produce final recommendation
| Verdict | Meaning |
|---|
| GO | Proceed with implementation — no blocking concerns |
| CAUTION | Proceed with specific mitigations — risks are manageable |
| STOP | Do not proceed — fundamental issues need resolution first |
STOP triggers (any one forces STOP): security auth bypass detected, architect finds design incompatibility, performance analysis shows exponential scaling, or devil’s advocate surfaces a false core assumption.
| Section | Content |
|---|
| Verdict | GO / CAUTION / STOP |
| Agreements | Points all personas aligned on |
| Conflicts & Resolutions | Table with each persona’s position + recommendation |
| Risk Summary | Risk, Severity, Mitigation |
| Recommendations | Numbered action items |
- Scenario — generate edge cases to test after Predict approves
- Plan — create implementation plan after debate
- Cook — implement after planning