🖥️
Best Viewed on Desktop
The technical documentation and app previews are optimized for desktop viewing.
← Back to Home🤖
LLM Agents
6 specialized agents executing sequentially. Each has a focused responsibility — complex reasoning decomposed into manageable steps.
Triage
Filter FPs
Correlator
Find links
AttackChain
Build chains
Exploit
PoC code
Remediation
Fix suggestions
Summary
Report
Configuration (Table 4)
| Agent | Model | Temp | Tokens | Responsibility |
|---|---|---|---|---|
| TriageAgent | gpt-4o | 0.1 | 1,000 | Filter FPs, adjust confidence |
| CorrelatorAgent | gpt-4o | 0.1 | 2,000 | Group by root cause |
| AttackChainAgent | gpt-4o | 0.3 | 3,000 | Build multi-step attack scenarios |
| ExploitGenerator | gpt-4o | 0.2 | 4,000 | Foundry-compatible PoC skeletons |
| RemediationAgent | gpt-4o | 0.1 | 2,000 | Fix recommendations + code patches |
| SummaryAgent | gpt-4o | 0.1 | 4,000 | Executive + technical report |
Agent Details
TriageAgent
In: All normalized findings + World
Out: Filtered findings + adjusted confidence
↳ Eliminates ~54% false positives
CorrelatorAgent
In: Triaged findings
Out: Findings grouped by root cause
↳ Reduces noise, identifies systemic problems
AttackChainAgent
In: Correlated findings + World
Out: Multi-step attack scenarios
↳ Compound vulnerabilities missed by individual tools
ExploitGenerator
In: High-severity findings + chains
Out: Foundry-compatible exploit skeletons
↳ PoC code — skeletons need refinement in 40%
RemediationAgent
In: Validated findings
Out: Fix recommendations + code patches
↳ Concrete remediations following best practices
SummaryAgent
In: All processed data
Out: Executive + technical report
↳ Actionable documentation for all stakeholders
⚠️ LLM Backend Migration Planned
Current: GPT-4o for all 6 agents. Planned: Gemini 3 (1M token context) enabling full contract suite analysis in a single pass.