AI-assisted security work is only useful when it produces verifiable findings without losing track of scope. The key idea here is to give an agent repeatable testing discipline—not just a collection of attack prompts—so it can help structure authorized assessments while leaving probing and proof to the operator.
What Sets It Apart
- Class-specific playbooks: Its skills cover web vulnerabilities and external enterprise attack surfaces, drawing on disclosed-report patterns to help turn broad security knowledge into testable hypotheses.
- Validation before reporting: A seven-question gate, evidence-hygiene guidance, and severity-aware reporting aim to filter out unsupported claims and reduce accidental exposure of sensitive data.
- Workflow beyond prompting: Reconnaissance, attack-surface prioritization, engagement scaffolding, and report preparation connect individual techniques into a more consistent assessment process.
- Portable knowledge, selective orchestration: The skill files can load in several agent harnesses, while slash commands and the engagement engine remain Claude Code-specific. This separates reusable guidance from harness-dependent automation.
Who It’s For
A good fit for security researchers and red teams conducting authorized, scope-bounded assessments who want an AI agent to follow a structured process. Look elsewhere if you need a standalone scanner, a general-purpose coding assistant, or guidance for internal Active Directory attacks, persistence, and post-exploitation; those are outside the project’s stated scope. The operator remains responsible for authorization, testing, and confirming evidence.