Review code across multiple security dimensions

Multi-dimension code review with adversarial verification

Multi-phase workflows requiring orchestrated agent handoffs and structured consensus.

• · pattern: string
• · baseRef: string

• · verdict: object (scores, winner, rationale)
• · report: markdown string
• · findings: array
• · confirmed: array

• · bash/CLI

• · Git repo available

• · Agents fail to produce mandatory adversarial refutation
• · Verification step catches regression, halts pipeline
• · Agent output fails schema validation

• · Structured agent output via JSON schema validation
• · Independent verification phase before acceptance
• · Frozen reference for regression detection