Agent KB_tests_p10 - grisuno/LazyOwn GitHub Wiki
Subsystem: tests (page 10 of 19)
Previous: KB_tests_p9.md
tests/test_improvements_spec.py
- Doc: Spec-driven validation of the three priority improvements.
- Layer: testing
- Language: py
- Symbols:
_FakeShell(class, line 77)class _FakeShell_FakeEngagement(class, line 90)class _FakeEngagement_FakeQueen(class, line 98)class _FakeQueen_FakeSwanResult(class, line 117)class _FakeSwanResult_FakeSwanOrchestrator(class, line 126)class _FakeSwanOrchestrator_FakeAdvisor(class, line 134)class _FakeAdvisor_build_test_orchestrator(function, line 142)def _build_test_orchestrator(tmp_root, daemon_available, hive_available, swan_available, daemon_payload)_UnavailableBackend(class, line 172)class _UnavailableBackendStatusBarConfigSpec(class, line 183)class StatusBarConfigSpec(TestCase)StatusBarSecuritySpec(class, line 205)class StatusBarSecuritySpec(TestCase)StatusBarRendererSpec(class, line 239)class StatusBarRendererSpec(TestCase)StatusBarManagerSpec(class, line 278)class StatusBarManagerSpec(TestCase)StatusBarSourceSpec(class, line 350)class StatusBarSourceSpec(TestCase)StatusBarFactorySpec(class, line 395)class StatusBarFactorySpec(TestCase)OrchestratorGoalValidationSpec(class, line 404)class OrchestratorGoalValidationSpec(TestCase)BackendRegistrySpec(class, line 444)class BackendRegistrySpec(TestCase)RouterPolicySpec(class, line 477)class RouterPolicySpec(TestCase)EventBusSpec(class, line 518)class EventBusSpec(TestCase)BackendAdapterSpec(class, line 550)class BackendAdapterSpec(TestCase)UnifiedOrchestratorSpec(class, line 601)class UnifiedOrchestratorSpec(TestCase)ConfigDedupeSpec(class, line 649)class ConfigDedupeSpec(TestCase)CommandSetActivationSpec(class, line 673)class CommandSetActivationSpec(TestCase)DocstringDisciplineSpec(class, line 706)class DocstringDisciplineSpec(TestCase)GraphifyAvailabilitySpec(class, line 745)class GraphifyAvailabilitySpec(TestCase)CmdIntegrationRegressionSpec(class, line 767)class CmdIntegrationRegressionSpec(TestCase)LiveShellBehaviourSpec(class, line 844)class LiveShellBehaviourSpec(TestCase)FactoryWiringSpec(class, line 996)class FactoryWiringSpec(TestCase)__init__(method, line 80)def __init__(self, custom_prompt)register_precmd_hook(method, line 86)def register_precmd_hook(self, hook)__init__(method, line 91)def __init__(self, payload)run(method, line 94)def run(self)__init__(method, line 99)def __init__(self, drones, summary)plan(method, line 103)def plan(self, goal, n_drones)dispatch(method, line 107)def dispatch(self, tasks)collect(method, line 110)def collect(self, ids)synthesize(method, line 113)def synthesize(self, results)__init__(method, line 118)def __init__(self, text, expert_id)__init__(method, line 127)def __init__(self, result)run(method, line 130)def run(self, task_type, goal, engagement_phase, timeout)__init__(method, line 135)def __init__(self, suggestions)suggest_next(method, line 138)def suggest_next(self, recent_commands, limit)__init__(method, line 173)def __init__(self, name)available(method, line 176)def available(self)run(method, line 179)def run(self, goal)test_config_is_immutable_dataclass(method, line 190)def test_config_is_immutable_dataclass(self)test_config_overrides_from_payload(method, line 196)def test_config_overrides_from_payload(self)test_config_ignores_invalid_overrides(method, line 200)def test_config_ignores_invalid_overrides(self)setUp(method, line 208)def setUp(self)tearDown(method, line 213)def tearDown(self)_reader(method, line 216)def _reader(self)test_reader_rejects_parent_traversal(method, line 219)def test_reader_rejects_parent_traversal(self)test_reader_rejects_absolute_paths(method, line 222)def test_reader_rejects_absolute_paths(self)test_reader_bounds_file_size(method, line 225)def test_reader_bounds_file_size(self)test_reader_returns_empty_on_missing_file(method, line 231)def test_reader_returns_empty_on_missing_file(self)test_reader_invalid_json_returns_none(method, line 234)def test_reader_invalid_json_returns_none(self)setUp(method, line 242)def setUp(self)test_render_plain_contains_all_fields(method, line 246)def test_render_plain_contains_all_fields(self)test_render_truncates_to_max_chars(method, line 252)def test_render_truncates_to_max_chars(self)test_render_strips_dangerous_substrings(method, line 258)def test_render_strips_dangerous_substrings(self)test_render_prompt_defaults_to_raw_ansi_for_prompt_toolkit(method, line 263)def test_render_prompt_defaults_to_raw_ansi_for_prompt_toolkit(self)test_render_prompt_opt_in_readline_mode_wraps_ansi_with_markers(method, line 270)def test_render_prompt_opt_in_readline_mode_wraps_ansi_with_markers(self)test_missing_source_key_raises(method, line 281)def test_missing_source_key_raises(self)test_collect_context_uses_fallbacks_on_failure(method, line 286)def test_collect_context_uses_fallbacks_on_failure(self)test_enabled_flag_accepts_strings_and_bools(method, line 304)def test_enabled_flag_accepts_strings_and_bools(self)test_install_registers_precmd_hook(method, line 325)def test_install_registers_precmd_hook(self)setUp(method, line 353)def setUp(self)tearDown(method, line 359)def tearDown(self)test_payload_target_picks_first_non_empty_key(method, line 362)def test_payload_target_picks_first_non_empty_key(self)test_payload_target_falls_back_to_default(method, line 366)def test_payload_target_falls_back_to_default(self)test_phase_prefers_world_model_over_payload(method, line 370)def test_phase_prefers_world_model_over_payload(self)test_finding_prefers_credentials_over_notes(method, line 375)def test_finding_prefers_credentials_over_notes(self)test_suggestion_uses_advisor_when_available(method, line 381)def test_suggestion_uses_advisor_when_available(self)test_suggestion_falls_back_when_advisor_raises(method, line 386)def test_suggestion_falls_back_when_advisor_raises(self)test_factory_returns_manager_with_four_sources(method, line 398)def test_factory_returns_manager_with_four_sources(self)setUp(method, line 407)def setUp(self)test_empty_goal_is_rejected(method, line 411)def test_empty_goal_is_rejected(self)test_unknown_mode_is_rejected(method, line 415)def test_unknown_mode_is_rejected(self)test_long_goal_is_bounded(method, line 419)def test_long_goal_is_bounded(self)test_defaults_are_applied(method, line 424)def test_defaults_are_applied(self)test_task_type_stays_empty_when_unset(method, line 429)def test_task_type_stays_empty_when_unset(self)test_numeric_fields_stay_zero_when_unset(method, line 433)def test_numeric_fields_stay_zero_when_unset(self)test_metadata_carries_request_id(method, line 438)def test_metadata_carries_request_id(self)test_duplicate_name_raises(method, line 447)def test_duplicate_name_raises(self)test_empty_name_raises(method, line 452)def test_empty_name_raises(self)test_order_is_preserved(method, line 465)def test_order_is_preserved(self)setUp(method, line 480)def setUp(self)test_explicit_mode_overrides_router(method, line 492)def test_explicit_mode_overrides_router(self)test_auto_picks_daemon_when_target_present(method, line 496)def test_auto_picks_daemon_when_target_present(self)test_auto_picks_hive_on_swarm_keyword(method, line 500)def test_auto_picks_hive_on_swarm_keyword(self)test_auto_picks_swan_on_exploit_keyword(method, line 504)def test_auto_picks_swan_on_exploit_keyword(self)test_unavailable_explicit_mode_returns_none(method, line 508)def test_unavailable_explicit_mode_returns_none(self)setUp(method, line 521)def setUp(self)tearDown(method, line 527)def tearDown(self)test_emit_appends_jsonline(method, line 530)def test_emit_appends_jsonline(self)test_emit_rejects_oversize_event(method, line 537)def test_emit_rejects_oversize_event(self)test_emit_rejects_non_serialisable(method, line 541)def test_emit_rejects_non_serialisable(self)test_file_permissions_are_restrictive(method, line 544)def test_file_permissions_are_restrictive(self)setUp(method, line 553)def setUp(self)tearDown(method, line 558)def tearDown(self)_goal(method, line 561)def _goal(self)test_daemon_requires_target(method, line 565)def test_daemon_requires_target(self)test_daemon_normalises_dict_result(method, line 570)def test_daemon_normalises_dict_result(self)test_hive_runs_full_lifecycle(method, line 577)def test_hive_runs_full_lifecycle(self)test_swan_normalises_result_object(method, line 584)def test_swan_normalises_result_object(self)test_backend_returns_error_when_engine_raises(method, line 590)def test_backend_returns_error_when_engine_raises(self)setUp(method, line 604)def setUp(self)tearDown(method, line 608)def tearDown(self)_events(method, line 611)def _events(self)test_validation_error_returns_invalid_result(method, line 617)def test_validation_error_returns_invalid_result(self)test_unavailable_backend_returns_unavailable_result(method, line 623)def test_unavailable_backend_returns_unavailable_result(self)test_auto_routes_with_target_to_daemon(method, line 628)def test_auto_routes_with_target_to_daemon(self)test_event_is_emitted_per_execution(method, line 634)def test_event_is_emitted_per_execution(self)test_result_to_dict_is_json_serialisable(method, line 642)def test_result_to_dict_is_json_serialisable(self)test_utils_has_no_class_config(method, line 658)def test_utils_has_no_class_config(self)test_core_config_is_canonical(method, line 664)def test_core_config_is_canonical(self)_discover(method, line 683)def _discover(self)test_orchestration_set_is_active(method, line 691)def test_orchestration_set_is_active(self)test_orchestration_set_exposes_required_verbs(method, line 697)def test_orchestration_set_exposes_required_verbs(self)test_orchestration_set_phase_and_category_set(method, line 701)def test_orchestration_set_phase_and_category_set(self)test_every_class_in_new_modules_has_docstring(method, line 709)def test_every_class_in_new_modules_has_docstring(self)test_modules_avoid_emoji_in_source(method, line 724)def test_modules_avoid_emoji_in_source(self)test_suggestion_fallback_when_factory_none(method, line 752)def test_suggestion_fallback_when_factory_none(self)test_graph_file_is_valid_json_when_present(method, line 758)def test_graph_file_is_valid_json_when_present(self)test_command_set_resolves_shell_via_cmd_property(method, line 781)def test_command_set_resolves_shell_via_cmd_property(self)test_status_bar_install_registers_precmd_hook_on_real_cmd2(method, line 795)def test_status_bar_install_registers_precmd_hook_on_real_cmd2(self)test_orchestration_commandset_finds_managers_on_bound_shell(method, line 818)def test_orchestration_commandset_finds_managers_on_bound_shell(self)test_default_prompt_join_is_newline(method, line 861)def test_default_prompt_join_is_newline(self)test_command_hint_source_returns_kill_chain_verb(method, line 866)def test_command_hint_source_returns_kill_chain_verb(self)test_command_hint_source_falls_back_when_provider_raises(method, line 892)def test_command_hint_source_falls_back_when_provider_raises(self)test_command_hints_helper_returns_phase_priority_verbs(method, line 913)def test_command_hints_helper_returns_phase_priority_verbs(self)test_daemon_falls_back_to_payload_rhost(method, line 921)def test_daemon_falls_back_to_payload_rhost(self)test_router_picks_daemon_when_only_payload_supplies_target(method, line 946)def test_router_picks_daemon_when_only_payload_supplies_target(self)test_router_ignores_unset_task_type_for_keyword_scoring(method, line 970)def test_router_ignores_unset_task_type_for_keyword_scoring(self)test_build_default_orchestrator_registers_three_backends(method, line 999)def test_build_default_orchestrator_registers_three_backends(self)test_build_default_manager_does_not_touch_filesystem(method, line 1003)def test_build_default_manager_does_not_touch_filesystem(self)_make_engine(function, line 154)def _make_engine(goal)_Broken(class, line 289)class _Broken_Static(class, line 308)class _Static_Static(class, line 329)class _Static_Broken(class, line 387)class _Broken_Unnamed(class, line 453)class _Unnamed_Boom(class, line 591)class _Boomcontains_emoji(function, line 730)def contains_emoji(text)_Shell(class, line 786)class _Shell(Cmd)_Shell(class, line 798)class _Shell(Cmd)_Shell(class, line 823)class _Shell(Cmd)broken(function, line 899)def broken()factory(function, line 931)def factory(goal)collect(method, line 290)def collect(self)__init__(method, line 309)def __init__(self, value)collect(method, line 312)def collect(self)__init__(method, line 330)def __init__(self, value)collect(method, line 333)def collect(self)suggest_next(method, line 388)def suggest_next(self)available(method, line 456)def available(self)run(method, line 459)def run(self, goal)run(method, line 592)def run(self)fake_hints(function, line 878)def fake_hints(last_cmd, phase, sessions_dir, limit)_Engine(class, line 934)class _Enginerun(method, line 935)def run(self_inner)
- Depends on:
cli/__init__.py,cli/commands/__init__.py,cli/commands/_dormancy.py,cli/commands/orchestration.py,cli/reactive_hints.py,cli/registry.py,cli/status_bar.py,core/config.py,skills/unified_orchestrator.py
tests/test_infra_disposable.py
- Doc: TDD unit tests for the disposable infrastructure contracts.
- Layer: testing
- Language: py
- Symbols:
test_parse_tunnel_urls_dedupes(function, line 16)def test_parse_tunnel_urls_dedupes()test_parse_tunnel_urls_empty(function, line 31)def test_parse_tunnel_urls_empty()test_parse_tunnel_urls_rejects_non_cloudflare(function, line 38)def test_parse_tunnel_urls_rejects_non_cloudflare()test_valid_providers_exact_allowlist(function, line 45)def test_valid_providers_exact_allowlist()test_infra_phases_as_c2(function, line 52)def test_infra_phases_as_c2()test_go_string_list_formats_slice(function, line 60)def test_go_string_list_formats_slice()test_resolve_fallback_urls_param_string(function, line 70)def test_resolve_fallback_urls_param_string()test_resolve_fallback_urls_param_list(function, line 78)def test_resolve_fallback_urls_param_list()test_implant_template_has_fallback_helpers(function, line 86)def test_implant_template_has_fallback_helpers()test_redirector_caddy_filters_paths(function, line 95)def test_redirector_caddy_filters_paths()test_redirector_compose_routes_to_caddy(function, line 103)def test_redirector_compose_routes_to_caddy()test_c2_compose_terminates_tls(function, line 111)def test_c2_compose_terminates_tls()test_terraform_firewall_restricts_c2_port(function, line 118)def test_terraform_firewall_restricts_c2_port()test_report_collects_history_and_loot(function, line 125)def test_report_collects_history_and_loot()test_report_template_summary_mentions_remediation_windows(function, line 137)def test_report_template_summary_mentions_remediation_windows()test_report_ai_failure_falls_back_to_template(function, line 150)def test_report_ai_failure_falls_back_to_template()test_gym_range_challenges_registered(function, line 161)def test_gym_range_challenges_registered()test_lab_range_profiles_registered(function, line 170)def test_lab_range_profiles_registered()test_range_compose_uses_valid_images(function, line 179)def test_range_compose_uses_valid_images()test_range_workstation_stays_alive(function, line 189)def test_range_workstation_stays_alive()test_range_dc_secret_wired(function, line 198)def test_range_dc_secret_wired()test_range_secret_generator_roundtrip(function, line 207)def test_range_secret_generator_roundtrip(tmp_path)test_parse_go_version_triples(function, line 217)def test_parse_go_version_triples()test_compile_commands_route_through_shell(function, line 228)def test_compile_commands_route_through_shell()test_garble_mismatch_detected(function, line 258)def test_garble_mismatch_detected()test_redirector_stale_threshold(function, line 272)def test_redirector_stale_threshold()test_format_timeline_event_summarizes_payload(function, line 279)def test_format_timeline_event_summarizes_payload()test_mitre_matrix_keeps_unknown_tactics(function, line 298)def test_mitre_matrix_keeps_unknown_tactics()test_range_backdoor_shell_published(function, line 309)def test_range_backdoor_shell_published()test_range_verify_confirms_root(function, line 318)def test_range_verify_confirms_root(capsys)test_range_verify_unknown_profile(function, line 336)def test_range_verify_unknown_profile(capsys)test_gym_range_next_steps(function, line 344)def test_gym_range_next_steps(tmp_path, monkeypatch)test_find_cloudflared_pids(function, line 357)def test_find_cloudflared_pids()test_compose_down_missing_file(function, line 373)def test_compose_down_missing_file()test_payload_schema_has_fallback_slot(function, line 380)def test_payload_schema_has_fallback_slot()
- Depends on:
cli/commands/infra.py,cli/commands/lab.py,core/payload_schema.py,modules/c2_builder.py,modules/professional_report.py,modules/redteam_gym.py,modules/session_cleanup.py,modules/world_model.py
tests/test_input_fuzz.py
- Doc: Fuzzing for input validators and rating boundaries.
- Layer: testing
- Language: py
- Symbols:
_random_host(function, line 80)def _random_host(rng)_host_inputs(function, line 85)def _host_inputs()_port_inputs(function, line 93)def _port_inputs()test_host_validators_never_raise(function, line 103)def test_host_validators_never_raise(value, capsys)test_accepted_hosts_have_no_shell_metacharacters(function, line 110)def test_accepted_hosts_have_no_shell_metacharacters(value, capsys)test_port_validator_never_raise_and_bounded(function, line 118)def test_port_validator_never_raise_and_bounded(value, capsys)test_known_good_values_accepted(function, line 126)def test_known_good_values_accepted(capsys)test_rate_plugin_rejects_non_integer_stars(function, line 139)def test_rate_plugin_rejects_non_integer_stars(tmp_path, stars)
- Depends on:
cli/plugin_tiers.py,core/validators.py
tests/test_intel_fusion.py
- Doc: Intel fusion + LLM defaults regression tests.
- Layer: testing
- Language: py
- Symbols:
_fixture_sessions(function, line 20)def _fixture_sessions(tmp_path)test_node_id_deterministic(function, line 43)def test_node_id_deterministic()test_build_fuses_all_sources(function, line 49)def test_build_fuses_all_sources(tmp_path)test_layouts_circular_spherical_force(function, line 58)def test_layouts_circular_spherical_force(tmp_path)test_summary_scores_and_suggests(function, line 69)def test_summary_scores_and_suggests(tmp_path)test_llm_defaults_seeded(function, line 78)def test_llm_defaults_seeded()test_factory_complete_rejects_empty(function, line 90)def test_factory_complete_rejects_empty()
- Depends on:
modules/intel_fusion.py,modules/lazy_loop.py,modules/llm_factory.py
tests/test_intelligence_engine.py
- Doc: Tests for IntelligenceEngine — collection, analysis, intelligence, dissemination.
- Layer: testing
- Language: py
- Symbols:
engine(function, line 19)def engine()nmap_xml(function, line 24)def nmap_xml(tmp_path)TestCollection(class, line 45)class TestCollectionTestAnalysis(class, line 108)class TestAnalysisTestIntelligenceProduction(class, line 159)class TestIntelligenceProductionTestCounterIntelligence(class, line 178)class TestCounterIntelligenceTestFullCycle(class, line 198)class TestFullCycleTestPlaceholderFiltering(class, line 215)class TestPlaceholderFilteringtest_collect_from_scan_services(method, line 46)def test_collect_from_scan_services(self, engine, nmap_xml)test_collect_from_scan_os(method, line 56)def test_collect_from_scan_os(self, engine, nmap_xml)test_collect_from_scan_domain(method, line 63)def test_collect_from_scan_domain(self, engine, nmap_xml)test_collect_from_scan_hosts(method, line 70)def test_collect_from_scan_hosts(self, engine, nmap_xml)test_collect_from_scan_missing_xml(method, line 76)def test_collect_from_scan_missing_xml(self, engine)test_collect_from_tool_parses_creds(method, line 81)def test_collect_from_tool_parses_creds(self, engine)test_collect_from_tool_filters_placeholders(method, line 86)def test_collect_from_tool_filters_placeholders(self, engine)test_collect_from_factstore(method, line 91)def test_collect_from_factstore(self, engine, tmp_path)test_analyze_produces_assessments(method, line 109)def test_analyze_produces_assessments(self, engine, nmap_xml)test_analyze_maps_apache_cve(method, line 115)def test_analyze_maps_apache_cve(self, engine)test_analyze_correlates_creds_to_hosts(method, line 131)def test_analyze_correlates_creds_to_hosts(self, engine)test_analyze_ranks_targets(method, line 143)def test_analyze_ranks_targets(self, engine)test_produce_intelligence_grades_assessments(method, line 160)def test_produce_intelligence_grades_assessments(self, engine)test_credential_exposure_detected(method, line 179)def test_credential_exposure_detected(self, engine)test_high_scan_volume_detected(method, line 188)def test_high_scan_volume_detected(self, engine)test_run_full_cycle_returns_summary(method, line 199)def test_run_full_cycle_returns_summary(self, engine, nmap_xml)test_get_intel_report_structured(method, line 205)def test_get_intel_report_structured(self, engine, nmap_xml)test_is_placeholder_detects_change_me(method, line 216)def test_is_placeholder_detects_change_me(self, engine)test_is_placeholder_rejects_real_values(method, line 220)def test_is_placeholder_rejects_real_values(self, engine)
- Depends on:
modules/intelligence_engine.py
tests/test_journal.py
- Doc: Tests for the GitHub Discussions engineering journal.
- Layer: testing
- Language: py
- Symbols:
FakeRunner(class, line 16)class FakeRunner_success_responses(function, line 34)def _success_responses()test_config_from_remote_slug(function, line 51)def test_config_from_remote_slug()test_config_rejects_bare_name(function, line 58)def test_config_rejects_bare_name()test_post_entry_uses_resolved_ids(function, line 64)def test_post_entry_uses_resolved_ids()test_entries_returns_nodes(function, line 73)def test_entries_returns_nodes()test_missing_category_raises(function, line 81)def test_missing_category_raises()test_graphql_errors_raise(function, line 93)def test_graphql_errors_raise()__init__(method, line 19)def __init__(self, responses)__call__(method, line 24)def __call__(self, args)
- Depends on:
scripts/journal.py
tests/test_killchain.py
- Doc: Tests for cli/killchain.py.
- Layer: testing
- Language: py
- Symbols:
_states(function, line 28)def _states(progress)test_empty_events_all_pending_without_world(function, line 32)def test_empty_events_all_pending_without_world()test_current_phase_from_step_events(function, line 38)def test_current_phase_from_step_events()test_phase_advance_event_sets_current(function, line 49)def test_phase_advance_event_sets_current()test_activity_and_reward_accumulate(function, line 62)def test_activity_and_reward_accumulate()test_world_phase_fallback_when_no_events(function, line 75)def test_world_phase_fallback_when_no_events()test_explicit_completed_phases_respected(function, line 81)def test_explicit_completed_phases_respected()test_unknown_phase_in_event_is_ignored(function, line 86)def test_unknown_phase_in_event_is_ignored()test_non_dict_payload_is_tolerated(function, line 92)def test_non_dict_payload_is_tolerated()
- Depends on:
cli/killchain.py
tests/test_killchain_auto_refresh.py
- Doc: BDD-style tests for the periodic kill-chain auto-refresh in the tips engine.
- Layer: testing
- Language: py
- Symbols:
_engine(function, line 12)def _engine()TestPeriodicAutoRefresh(class, line 22)class TestPeriodicAutoRefreshtest_shows_immediately_on_phase_change(method, line 23)def test_shows_immediately_on_phase_change(self)test_shows_on_cadence_without_phase_change(method, line 31)def test_shows_on_cadence_without_phase_change(self)test_cadence_below_every_surpresses(method, line 41)def test_cadence_below_every_surpresses(self)test_phase_change_beats_periodic_cadence(method, line 50)def test_phase_change_beats_periodic_cadence(self)test_disabled_engine_never_shows(method, line 59)def test_disabled_engine_never_shows(self)test_zero_every_and_no_phase_change_never_shows(method, line 67)def test_zero_every_and_no_phase_change_never_shows(self)
- Depends on:
cli/tips_engine.py
tests/test_killchain_gap_signal.py
- Doc: Tests for :class:
cli.recommendation_signals.KillchainGapSignal. - Layer: testing
- Language: py
- Symbols:
sessions_dir(function, line 23)def sessions_dir()_write_world_model(function, line 28)def _write_world_model(sessions_dir, data)TestKillchainGapSignalConstruction(class, line 32)class TestKillchainGapSignalConstructionTestGapExploitedNoPrivesc(class, line 38)class TestGapExploitedNoPrivescTestGapOwnedNoCreds(class, line 82)class TestGapOwnedNoCredsTestGapScanNoEnum(class, line 110)class TestGapScanNoEnumTestGapCredsNoLateral(class, line 141)class TestGapCredsNoLateralTestKillchainGapSignalIntegration(class, line 169)class TestKillchainGapSignalIntegrationtest_name_is_gap_source(method, line 33)def test_name_is_gap_source(self)test_exploited_linux_recommends_linpeas(method, line 39)def test_exploited_linux_recommends_linpeas(self, sessions_dir)test_exploited_windows_recommends_winpeas(method, line 51)def test_exploited_windows_recommends_winpeas(self, sessions_dir)test_unscanned_host_no_proposals(method, line 63)def test_unscanned_host_no_proposals(self, sessions_dir)test_no_world_model_returns_empty(method, line 75)def test_no_world_model_returns_empty(self, sessions_dir)test_owned_no_credentials_recommends_lazydump(method, line 83)def test_owned_no_credentials_recommends_lazydump(self, sessions_dir)test_owned_with_credentials_no_proposals(method, line 96)def test_owned_with_credentials_no_proposals(self, sessions_dir)test_scanned_no_enum_recommends_gobuster(method, line 111)def test_scanned_no_enum_recommends_gobuster(self, sessions_dir)test_scanned_with_enum_recent_no_proposals(method, line 123)def test_scanned_with_enum_recent_no_proposals(self, sessions_dir)test_credentials_no_lateral_recommends_crackmapexec(method, line 142)def test_credentials_no_lateral_recommends_crackmapexec(self, sessions_dir)test_no_credentials_no_lateral_proposals(method, line 155)def test_no_credentials_no_lateral_proposals(self, sessions_dir)test_multiple_gaps_detected_simultaneously(method, line 170)def test_multiple_gaps_detected_simultaneously(self, sessions_dir)
- Depends on:
cli/recommendation.py,cli/recommendation_signals.py
tests/test_killchain_snapshot.py
- Doc: BDD-style tests for the unified kill-chain snapshot contract.
- Layer: testing
- Language: py
- Symbols:
_seed_snapshot(function, line 17)def _seed_snapshot(path, phase, completed, current)_states(function, line 37)def _states(snapshot)_encrypt_at(function, line 41)def _encrypt_at(path, password, salt)TestKillChainSnapshot(class, line 52)class TestKillChainSnapshotTestEncryptedStateTransparency(class, line 102)class TestEncryptedStateTransparencytest_empty_snapshot_defaults_to_recon(method, line 53)def test_empty_snapshot_defaults_to_recon(self, tmp_path)test_snapshot_reflects_explicit_phase_and_completed(method, line 64)def test_snapshot_reflects_explicit_phase_and_completed(self, tmp_path)test_snapshot_is_json_serialisable(method, line 80)def test_snapshot_is_json_serialisable(self, tmp_path)test_advance_then_snapshot_is_consistent(method, line 90)def test_advance_then_snapshot_is_consistent(self, tmp_path)test_read_state_dict_decrypts_at_rest(method, line 103)def test_read_state_dict_decrypts_at_rest(self, tmp_path, monkeypatch)test_missing_file_reads_empty(method, line 123)def test_missing_file_reads_empty(self, tmp_path)test_encrypted_without_password_reads_empty(method, line 131)def test_encrypted_without_password_reads_empty(self, tmp_path, monkeypatch)
- Depends on:
modules/killchain.py,modules/world_model.py
tests/test_killchain_unified.py
- Doc: Tests for unified killchain phase: read_phase, write_phase, phase mapping helpers.
- Layer: testing
- Language: py
- Symbols:
TestPhaseMapping(class, line 19)class TestPhaseMappingTestWritePhaseWithWorldModel(class, line 47)class TestWritePhaseWithWorldModeltest_engagement_phase_to_cli_maps_all(method, line 20)def test_engagement_phase_to_cli_maps_all(self)test_cli_phase_to_host_state_maps_correctly(method, line 29)def test_cli_phase_to_host_state_maps_correctly(self)test_phase_rank_returns_correct_index(method, line 40)def test_phase_rank_returns_correct_index(self)sessions_dir(method, line 51)def sessions_dir(self)test_write_phase_advances_hosts(method, line 79)def test_write_phase_advances_hosts(self, sessions_dir)test_write_phase_invalid_returns_false(method, line 95)def test_write_phase_invalid_returns_false(self, sessions_dir)test_write_phase_completed_phases_tracks_progress(method, line 99)def test_write_phase_completed_phases_tracks_progress(self, sessions_dir)
- Depends on:
cli/ops_commands.py,modules/world_model.py
tests/test_killchain_unified_v2.py
- Doc: Comprehensive tests for modules.killchain — the unified single source of truth.
- Layer: testing
- Language: py
- Symbols:
TestKillChainConfig(class, line 31)class TestKillChainConfigTestKillChainCurrentPhase(class, line 103)class TestKillChainCurrentPhaseTestKillChainAdvancePhase(class, line 184)class TestKillChainAdvancePhaseTestKillChainGetProgress(class, line 295)class TestKillChainGetProgressTestKillChainHelpers(class, line 363)class TestKillChainHelpersTestPhaseStatusDataclass(class, line 410)class TestPhaseStatusDataclasstest_phases_is_8_tuple(method, line 34)def test_phases_is_8_tuple(self)test_phases_are_in_correct_kill_chain_order(method, line 40)def test_phases_are_in_correct_kill_chain_order(self)test_compact_phases_are_in_correct_order(method, line 44)def test_compact_phases_are_in_correct_order(self)test_all_phases_have_labels(method, line 48)def test_all_phases_have_labels(self)test_all_phases_have_colors(method, line 53)def test_all_phases_have_colors(self)test_all_phases_have_rich_colors(method, line 58)def test_all_phases_have_rich_colors(self)test_engagement_to_cli_covers_all_engagement_phases(method, line 62)def test_engagement_to_cli_covers_all_engagement_phases(self)test_cli_to_host_state_returns_expected(method, line 70)def test_cli_to_host_state_returns_expected(self)test_phase_index_valid_and_invalid(method, line 80)def test_phase_index_valid_and_invalid(self)test_is_valid_phase(method, line 86)def test_is_valid_phase(self)test_compact_phases_and_labels(method, line 91)def test_compact_phases_and_labels(self)test_world_model_path(method, line 96)def test_world_model_path(self)_reset_wm_singleton(method, line 107)def _reset_wm_singleton(self)test_returns_recon_when_no_world_model_exists(method, line 118)def test_returns_recon_when_no_world_model_exists(self)test_reads_from_world_model_host_state(method, line 125)def test_reads_from_world_model_host_state(self)test_raw_json_override_wins_when_higher_rank(method, line 140)def test_raw_json_override_wins_when_higher_rank(self)test_raw_json_override_ignored_when_lower_rank(method, line 156)def test_raw_json_override_ignored_when_lower_rank(self)test_falls_back_to_legacy_phase_key(method, line 172)def test_falls_back_to_legacy_phase_key(self)_reset_wm_singleton(method, line 188)def _reset_wm_singleton(self)test_advance_writes_current_phase_and_phase_keys(method, line 199)def test_advance_writes_current_phase_and_phase_keys(self)test_advance_tracks_completed_phases(method, line 215)def test_advance_tracks_completed_phases(self)test_advance_invalid_phase_returns_false(method, line 233)def test_advance_invalid_phase_returns_false(self)test_advance_advances_world_model_hosts(method, line 243)def test_advance_advances_world_model_hosts(self)test_advance_with_host_does_not_touch_unrelated_hosts(method, line 260)def test_advance_with_host_does_not_touch_unrelated_hosts(self)test_advance_does_not_downgrade_cached_world_model_state(method, line 276)def test_advance_does_not_downgrade_cached_world_model_state(self)_reset_wm_singleton(method, line 299)def _reset_wm_singleton(self)test_all_pending_when_nothing_done(method, line 310)def test_all_pending_when_nothing_done(self)test_progress_reflects_completed_and_active(method, line 323)def test_progress_reflects_completed_and_active(self)test_progress_has_colors_and_labels(method, line 347)def test_progress_has_colors_and_labels(self)test_phases_for_display_returns_triples(method, line 366)def test_phases_for_display_returns_triples(self)test_compact_progress_returns_string(method, line 376)def test_compact_progress_returns_string(self)test_engagement_phase_to_cli_maps_all(method, line 382)def test_engagement_phase_to_cli_maps_all(self)test_cli_phase_to_host_state_maps_all(method, line 391)def test_cli_phase_to_host_state_maps_all(self)test_get_killchain_returns_class(method, line 399)def test_get_killchain_returns_class(self)test_phase_index_returns_correct(method, line 403)def test_phase_index_returns_correct(self)test_phase_status_is_immutable(method, line 413)def test_phase_status_is_immutable(self)test_phase_status_fields_match_config(method, line 418)def test_phase_status_fields_match_config(self)
- Depends on:
modules/killchain.py,modules/world_model.py
tests/test_lab_target_visibility.py
- Doc: Lab target visibility: start/status always show the container IP.
- Layer: testing
- Language: py
- Symbols:
test_start_on_running_lab_reports_ip(function, line 14)def test_start_on_running_lab_reports_ip(capsys, monkeypatch)test_status_reports_ip(function, line 27)def test_status_reports_ip(capsys, monkeypatch)
- Depends on:
cli/commands/lab.py
tests/test_lazy_agent.py
- Doc: Agentic tool-calling: worker agent path, resume and main sessions.
- Layer: testing
- Language: py
- Symbols:
_tool_call(function, line 13)def _tool_call(tool_id, name, arguments)_message(function, line 19)def _message(content, tool_calls)_Completions(class, line 23)class _Completionsfake_model(function, line 37)def fake_model(script)_sandbox(function, line 43)def _sandbox(tmp_path, monkeypatch)_worker(function, line 50)def _worker(loop, engine, task_id, graph)test_worker_agentic_tool_calling(function, line 63)def test_worker_agentic_tool_calling(tmp_path, monkeypatch)test_worker_agentic_resume_seeds_prior_turns(function, line 104)def test_worker_agentic_resume_seeds_prior_turns(tmp_path, monkeypatch)test_main_session_persists_and_resumes(function, line 138)def test_main_session_persists_and_resumes(tmp_path, monkeypatch)test_model_keepalive_tick_while_waiting(function, line 152)def test_model_keepalive_tick_while_waiting(tmp_path, monkeypatch)test_tool_calling_flag_honors_payload(function, line 175)def test_tool_calling_flag_honors_payload(tmp_path, monkeypatch)test_external_opencode_degrades_without_binary(function, line 184)def test_external_opencode_degrades_without_binary(monkeypatch)test_lazy_agent_command_session_lifecycle(function, line 194)def test_lazy_agent_command_session_lifecycle(tmp_path, monkeypatch)__init__(method, line 26)def __init__(self, script)create(method, line 30)def create(self)_executor(function, line 85)def _executor(command)_Completions(class, line 162)class _Completionscreate(method, line 163)def create(self)
- Depends on:
cli/commands/lazyops.py,core/config.py,modules/__init__.py,modules/lazy_agent.py,modules/lazy_engine.py,modules/lazy_loop.py
tests/test_lazy_bridge_traffic.py
- Doc: Incremental LAZY traffic mirroring (stall regression).
- Layer: testing
- Language: py
- Symbols:
_sandbox(function, line 16)def _sandbox(tmp_path, monkeypatch)_append(function, line 29)def _append(path, rows)_exchanges(function, line 35)def _exchanges(sessions_dir)test_traffic_mirror_is_incremental(function, line 39)def test_traffic_mirror_is_incremental(tmp_path, monkeypatch)test_partial_line_is_retried(function, line 61)def test_partial_line_is_retried(tmp_path, monkeypatch)test_progress_callback_survives_failures(function, line 78)def test_progress_callback_survives_failures(tmp_path, monkeypatch)_boom(function, line 83)def _boom(_text)
- Depends on:
modules/__init__.py,modules/traffic_recorder.py
Next: KB_tests_p11.md