bluechips-zhao/diting is an LLM-based autonomous penetration testing workbench built with Go, Next.js and PostgreSQL. The latest release published in the repository is v2.7.0. However, the provided materials list no changes for this version, so newly added or modified features cannot be identified. The repository description and README introduce the asset graph, multi-agent collaboration, traffic evidence and human approval features.
The tool plans and executes penetration testing tasks using a language model, then manages findings and evidence into reports. Agents are split into planning, execution, conversation and reporting roles, and are given token and tool-use budgets. Vulnerabilities and evidence can be bundled into versions for re-testing or report regeneration, and the code also includes a feature that links requests and responses as evidence to export HAR files and raw packets. There are also rules requiring human approval before tool calls. It is based on ARTEX and is licensed under AGPL-3.0.
That said, the materials explain that the environments actually validated are limited to Windows amd64 and Linux amd64 under Docker Desktop, and that other platforms are unvalidated. The README advises using it only against targets you own or have explicit permission to test. Therefore the multi-agent and evidence management configuration can serve as a reference, but no basis is given for concluding that it behaves the same way in other environments.