Autonomous OODA-loop pentesting agent. Hermes brain, Kali + Playwright tools over MCP, ATT&CK as the Orient language.
Python ai-agentsautonomous-agentskali-linuxllm-agentsoffensive-securityooda-looppentestingplaywrightsecurity
Plugin fixtures for test_plugin_loader
Example plugin tree that test_plugin_loader.py points at as plugins_root. Each
subdirectory exercises one loader path:
aaa_first/zzz_last: valid plugins (check deterministic alphabetical order).sibling_import:plugin.pythat doesimport helper(sibling) — demonstrates MEDIUM-3.broken_import.src→ renamed in the test to a dir with a broken import (see note). Instead we usebroken_importwith a nonexistentimport(phase 'import').no_api: withoutPLUGIN_API(phase 'api').bad_api: incompatiblePLUGIN_API(phase 'api').no_register: without a callableregister(phase 'register-missing').raises_register:registerthat registers a tool and then raises (phase 'register' + NIT-8).example.disabled: disabled dir (ignored).not_a_plugin: dir withoutplugin.py(skipped).
Invalid directory names (MEDIUM-4) CANNOT exist as a portable folder on all FS, so that case
is created on the fly inside the test with tmp_path.