Specialized agents

Orchestrated intelligence, not a single model.

HELIX isn't one giant prompt pretending to be an expert at everything. It's a team of specialists, each a focused agent with its own toolset, coordinated by a planner and sharing what they learn.

How the team operates

The planner generates an attack plan in buckets and routes each bucket to a specialized sub-agent with its own offensive toolset. Findings flow through a shared blackboard, so a discovery made by one agent informs the next, recon feeds injection, injection feeds chaining, and a skeptic checks every claim before it ships.

Web

Recon, authentication, access control, injection, business logic and chain hunting.

VANGUARD

Web recon

Maps before you wake. Enumerates surface, endpoints and entry points to seed the engagement.

reconenumeration
VENOM

Injection

Payload alchemist. Probes SQLi, XSS, template and command injection with real tooling.

injectionpayloads
KEYMASTER

Authentication

Takes keys, never doors. Attacks login, session and token flows for auth weaknesses.

authsessions
ASCENT

Access control / IDOR

No role too high. Hunts broken access control and insecure direct object references.

access controlIDOR
CASCADE

Chain hunter

Sees bugs as dominoes. Stitches individual weaknesses into a higher-impact path.

chainingimpact
LOOPHOLE

Business logic

Abuses the rules of the application itself, workflows, limits and assumptions that code never enforced.

business logicworkflow abuse

API

Spec-driven hunting for object- and function-level authorization flaws.

BLUEPRINT

Spec ingestion

Your spec is my map. Ingests OpenAPI, Swagger and GraphQL to model every route and parameter.

OpenAPIGraphQL
CLEARANCE

BOLA / BFLA

Tests object- and function-level authorization, the access-control flaws that dominate API breaches.

BOLABFLA
OVERFLOW

Mass assignment

"You forgot to strip a field." Pushes unexpected fields through endpoints to seize properties the API never meant to expose.

mass assignmentfield tampering
MARIONETTE

Mobile runtime

Your app dances now. Drives Frida and Objection to instrument iOS and Android binaries at runtime.

FridaObjection
ATLAS

Cloud

Maps AWS and Kubernetes exposure and probes IAM privilege escalation, exposed storage and metadata SSRF.

AWSKubernetes
ORACLE

AI & LLM

Asks the model what it shouldn't tell. Drives prompt injection, system-prompt leakage and RAG poisoning.

prompt injectionOWASP LLM

Coordination

The agents that keep the team honest, organized and reproducible.

CONDUCTOR

Orchestration

Plans the engagement in buckets and routes each one to the right specialist.

planningrouting
DOUBT

Skeptic

Every claim needs a witness. Refutes anything that lacks runtime corroboration.

verificationfalse-positive
PROOF

Reproducibility

Turns confirmed findings into copy-pasteable reproducers anyone can replay.

reproducerevidence
QUILL

Reporting

Writes the triaged report, CVSS, CWE and language-specific remediation per finding.

CVSS / CWEremediation
40+
specialized agents across web, API, mobile, cloud, AI and coordination
~100
real offensive tools the agents wield
1
shared blackboard every finding flows through

Put the whole team on your target

One planner, dozens of specialists, one triaged report.