Awesome AI Agents › Red-Teaming Platforms

SanMuzZzZz/LuaN1aoAgent

⭐ 1308 Python repository created 2025-12-04

LuaN1aoAgent v2 is a complete TypeScript rewrite of an earlier Python agent for authorised security research, built on the Pi SDK with explicit agent boundaries, durable events and artifacts, and observable tool actions. Three roles divide the work. The Planner maintains an evolving task graph instead of a linear checklist, reading compact task, reasoning and operation graph views, creating or patching goal-level tasks rather than prescribing low-level actions, and controlling dependencies, priority, concurrency of independent tasks, scope and per-task budgets; it reconciles ready tasks against available capacity after each graph change instead of waiting for a whole parallel wave, and commits decisions through a structured planner_submit tool. The Executor receives a bounded task envelope, chooses its own tool strategy, and records public intent, tool input and output, usage, errors and final results, preserving large outputs as immutable artifacts so agent context does not inflate; it reuses one persisted session lineage and workspace across epochs of a task while different tasks stay isolated. The Observer runs in two independent modes: a hot-path Supervisor that inspects recent executor actions and decides whether to continue, checkpoint, stop or hand control back to the Planner, and an asynchronous Projector that converts normalised observations into reasoning and operation graph deltas without blocking execution, each invocation using a fresh session with no hidden shared history. The causal reasoning graph links evidence to hypotheses, confirmed vulnerabilities and successful exploits, keeps uncertainty explicit, and enforces provenance so confirmed vulnerability and successful exploit nodes cannot be written without evidence references. A workbench UI shows live traces and the reasoning graph. The README states that v1 benchmark results do not carry over and will be republished only after reproducible reruns.

https://github.com/SanMuzZzZz/LuaN1aoAgent

securitypenetration-testingautonomous-agentknowledge-graphtypescript