guardian-cli

Why we included this project

Pen testers who operate with explicit authorization will appreciate Guardian as a command-line orchestrator that puts an AI planning layer on top of the tools they already use. It takes established scanners and fuzzers such as nuclei, sqlmap, nmap, and BloodHound and arranges them into step-by-step workflows, with a language model deciding which tool to invoke next and reviewing each result before moving on. The multi-agent design splits the work among planner, tool selector, analyst, and reporter roles, and adds a red/blue/judge debate to sort out ambiguous findings, which helps cut false positives and keeps the reasoning open to inspection. Every finding links back to the exact command and raw output, and results can be exported to SARIF, DefectDojo, or Slack. Guardian is built for legal, authorized testing, with scope validation, confirmation gates for intrusive actions, and a log of every AI decision to keep the process accountable.