Audits

An audit is one offensive engagement against a target domain. You choose a mode, pick targets, and Golem takes over.

The three modes

Shallow

OWASP Top 10 + PoC only. ~30 min. No post-exploitation.

Deep

Full 10-phase attack chain. ~2 hrs. Includes post-exploitation.

Autonomous

Operator-defined mission brief. ~4 hrs. You write the scope.

How an engagement runs

You create the audit                  →  POST /api/workspaces/{id}/audits
Golem AI provisions a darkops sandbox →  hardened Ubuntu desktop with full toolkit
Golem loads its mission brief       →  determined by mode + your scope
The agent executes the attack chain   →  recon → enum → vuln scan → exploit → post-ex
Findings stream in continuously       →  appended to the audit every 5 seconds
Engagement completes                  →  PDF + asciinema + score generated

You can watch the entire engagement in real time from the dashboard, or wait for the completion notification (email or Slack).

What every audit produces

Findings — confirmed vulnerabilities with severity, evidence, and remediation. See Findings & Evidence.
Security score — letter grade A+ to F derived from finding weights. See Security Score.
PDF report — cover page, executive summary, per-finding pages with evidence and recommendations. See Reports & Recordings.
Asciinema recording — full terminal session replay of every command Golem ran.
Screenshots — automatic visual evidence captured at the moment of discovery.

Choosing a mode

If you want to…	Use
Gate every release in CI	Shallow — fast enough to run on every PR
Replace a quarterly pentest	Deep — full chain including post-exploitation
Simulate a specific threat actor	Autonomous — write the brief like you’d brief a human
Verify a specific finding is remediated	Shallow with a narrow target list
Run a red-team simulation	Autonomous with assumed-breach scope

Audit statuses

Status	Description
`queued`	Sandbox provisioning in progress
`running`	Golem actively engaging the target
`completed`	Engagement finished, report ready
`failed`	Run encountered an unrecoverable error — partial findings preserved

Time budgets

Each mode has a default time budget that controls when Golem stops exploring and starts writing the report:

Shallow: 30 minutes
Deep: 2 hours
Autonomous: 4 hours

The agent’s final phase is always “write report” — even if interrupted, you’ll receive whatever findings were confirmed before the budget expired.

Authorization

Golem AI assumes every target you submit is authorized for full-scope offensive testing. You are responsible for ensuring written authorization before each engagement.

Getting Started

Audits

How It Works

Platform

Billing

Overview

Audits

The three modes

Shallow

Deep

Autonomous

How an engagement runs

What every audit produces

Choosing a mode

Audit statuses

Time budgets

Authorization

​Audits

​The three modes

Shallow

Deep

Autonomous

​How an engagement runs

​What every audit produces

​Choosing a mode

​Audit statuses

​Time budgets

​Authorization

Audits

The three modes

How an engagement runs

What every audit produces

Choosing a mode

Audit statuses

Time budgets

Authorization