Reconator v2: AI-Powered Reconnaissance with Adaptive Pipeline Orchestration

Reconator v2, an AI-powered reconnaissance framework that introduces adaptive pipeline orchestration for attack surface discovery. Unlike prior work that executes a fixed sequence of security tools, Reconator v2 employs a DAG-based execution model with two-stage AI checkpoints that analyze intermediate results and dynamically adjust downstream scanning strategies. Key innovations include: (1) a directed acyclic graph pipeline with 17 phases and soft dependencies enabling maximum parallelization, (2) mid-pipeline AI checkpoints that prioritize targets and generate adaptive wordlists based on discovered infrastructure, (3) TOON (Tree-Oriented Object Notation), a token-efficient data encoding for LLM prompts, (4) technology-aware vulnerability scanning, (5) passive OSINT intelligence module, (6) Graph-of-Thought vulnerability chain analysis with interactive visualization (Attack Paths), (7) automated vulnerability post-processing, and (8) a self-contained HTML report (9) Interactive Web Dashboard. Author's Note This preprint documents Reconator v2, the Go-based rewrite with AI orchestration, evolving from the bash-based Reconator v1 framework introduced in our 2021 preprint (DOI: 10.5281/zenodo.19979528). The v2 implementation introduces a web dashboard, distributed scanning fleet, and LLM-driven adaptive orchestration. The Reconator project was first open-sourced on GitHub in November 2020 and is continuously maintained at github.com/rootsploit/reconator under AGPL-3.0-or-later. Notes Originally drafted January 2026 (announced via Twitter on 2026-01-08). Deposited to Zenodo in May 2026. Companion preprint to the 2021 Reconator v1 paper. License: AGPL-3.0-or-later (code), CC-BY-4.0 (this paper).