AI agents

Ten specialized advisory roles share one authorization state. AI cannot override the server’s safety gate.

One orchestrator. Two independent safety gates.

Scope Guard and Compliance enforce deterministic blocks in server logic. Other agent roles produce AI-generated advice or inspect supplied evidence. No agent performs external testing or autonomous disclosure.

Centralized policy stateAI-generated, not live-searchedSafety rules enforced

Advisory role

Policy Agent

Interpret supplied program policies without inventing authorization.

1 recorded activities

Enforcement + advisory

Scope Guard Agent

Block ambiguous, expired, or non-allowlisted assets deterministically.

1 recorded activities

Advisory role

Recon Agent

Summarize supplied reconnaissance; external discovery is unavailable.

1 recorded activities

Advisory role

Bug Hunter Agent

Inspect sanitized response headers with four deterministic modules.

1 recorded activities

Advisory role

Validation Agent

Validate capture consistency; require humans to verify impact.

1 recorded activities

Advisory role

False Positive Agent

Identify missing context and unsupported exploitability claims.

1 recorded activities

Advisory role

Severity Agent

Assess severity from evidence and explicit impact assumptions.

1 recorded activities

Advisory role

Report Writer Agent

Draft evidence-backed reports for independent human review.

1 recorded activities

Enforcement + advisory

Compliance Agent

Enforce fixed safety rules and recommend blocking unsafe operations.

1 recorded activities

Advisory role

Orchestrator Agent

Coordinate the advisory workflow around the central policy state.

1 recorded activities