The brain agent and its sub-agents
AgentRidge is not one monolithic model answering prompts. It is a team of agents with a clear chain of command: one brain that thinks about the engagement as a whole, and specialists that do the focused work.
The Agent Ridge Principal (the brain)
The Principal is the orchestrator. It holds the mission's objective, builds and revises the plan, decides which specialist to dispatch next, judges whether a result is good enough, and declares when the engagement's definition of done has been reached.
What the brain is responsible for:
- Planning — turning a mission brief into an ordered set of goals.
- Delegation — handing each goal to the right sub-agent.
- Supervision — reading what came back, deciding the next move, and reopening earlier phases when new surface appears.
- Persistence — never stopping at "possible vulnerability"; pushing every lead until it is either proven or exhausted.
- Judgement — keeping the work in scope and proportionate to the chosen engagement mode.
The Ridge Sub-Agents (the specialists)
The Principal delegates to Ridge Sub-Agents, each tuned for one kind of job:
| Sub-agent role | What it does |
|---|---|
| Recon & enumeration | Discovers hosts, ports, services, web and API surface, identity endpoints, and prior leads worth re-testing. |
| Exploitation & verification | Generates and runs the actual test commands, then confirms whether a vulnerability is real with a live proof of concept. |
| Parsing & reasoning | Reads raw tool output and messy responses and distills them into structured, meaningful facts. |
Sub-agents are workers, not decision-makers. They report results — and their own failures — back up to the Principal, which decides what happens next.
How a single step flows
Graceful fallback
Autonomy means the system keeps going even when a part of it stalls. If a sub-agent times out, exhausts its budget, or the local reasoning step fails, the Principal does not end the engagement — it continues the work itself using the available tooling and pushes each remaining lead further. The result is a run that keeps making progress instead of giving up at the first obstacle.
Why split the work this way?
- Depth without losing the thread. Specialists can go deep on one task while the brain keeps the whole objective in view.
- Resilience. One stuck specialist doesn't stall the mission.
- Better evidence. A dedicated parsing role means findings arrive as clean, structured facts — ready for the vulnerability register and the report.