RedTeam — Pentest Operator Persona

RedTeam is a complete OpenClaw persona package that turns an AI agent into a structured offensive security operator. Not an assistant that helps with pentesting — an operator that runs engagements.

7 interconnected persona files, a 9-phase methodology skill, and a persistent knowledge graph that compounds across every engagement. The methodology draws from DARPA AIxCC Cyber Reasoning System competition analysis and OSCP (PEN-200) training frameworks.

For operators who already know what SQL injection is. RedTeam handles the systematic work — enumeration, state tracking, credential management, finding documentation, attack path reasoning — so you can focus on the judgment calls.

Key features: • 9-phase pentest methodology from engagement setup through final report delivery • Attack path reasoning based on expected value — probability × impact, weighed against time cost • Park-and-return registry that automatically retests stalled paths when new credentials surface • Persistent knowledge graph (PARA-structured) that compounds technique effectiveness, client patterns, and tool knowledge across engagements • Professional documentation discipline with observational language, evidence chains, and severity calibration • Three-tier AI operations (augmented, opinionated, driven) with validation at every stage • Engagement state management with time-based strategy shifts and heartbeat monitoring • OSCP-aligned tooling and methodology

Includes: SOUL.md, IDENTITY.md, AGENTS.md, TOOLS.md, MEMORY.md, HEARTBEAT.md, USER.md, 9-phase methodology skill, nightly extraction skill, and a full knowledge graph with templates for engagements, targets, techniques, clients, tools, remediation, wordlists, and payloads.

Not the most technically capable AI pentest tool available. The most methodologically structured one. Trades raw automation for operator judgment, engagement discipline, and cross-engagement learning.