android-agent-harness 1.3.2__tar.gz → 1.3.4__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (279) hide show
  1. android_agent_harness-1.3.4/PKG-INFO +455 -0
  2. android_agent_harness-1.3.4/README.md +426 -0
  3. android_agent_harness-1.3.4/agents/VERSION +1 -0
  4. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/release_checksums.json +17 -17
  5. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_ecosystem_upgrades_selftest.py +91 -0
  6. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_project_intelligence_selftest.py +163 -0
  7. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_workflow/planning.py +5 -1
  8. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/change_classifier.py +24 -1
  9. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/delegation_policy.py +74 -9
  10. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/doctor/engine.py +23 -1
  11. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/generate_project_context.py +20 -2
  12. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/phase_review.py +28 -2
  13. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/plan_document.py +8 -0
  14. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/pre_invocation_reminder.py +20 -1
  15. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/project_context.py +113 -2
  16. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/project_graph.py +35 -4
  17. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/review_package.py +3 -2
  18. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/roborazzi_gate.py +79 -32
  19. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/session_forensics.py +2 -0
  20. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/task_context.py +28 -16
  21. android_agent_harness-1.3.4/android_agent_harness.egg-info/PKG-INFO +455 -0
  22. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/pyproject.toml +1 -1
  23. android_agent_harness-1.3.2/PKG-INFO +0 -737
  24. android_agent_harness-1.3.2/README.md +0 -708
  25. android_agent_harness-1.3.2/agents/VERSION +0 -1
  26. android_agent_harness-1.3.2/android_agent_harness.egg-info/PKG-INFO +0 -737
  27. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/LICENSE +0 -0
  28. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/.gitignore +0 -0
  29. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/check-strings.md.template +0 -0
  30. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/commit-msg.md.template +0 -0
  31. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/crash-triage.md.template +0 -0
  32. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/debug.md.template +0 -0
  33. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/deliver.md.template +0 -0
  34. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/doctor.md.template +0 -0
  35. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/new-feature.md.template +0 -0
  36. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/perf-audit.md.template +0 -0
  37. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/pm-tracker.md.template +0 -0
  38. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/preflight.md.template +0 -0
  39. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/command-packs/test-quality-audit.md.template +0 -0
  40. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/harness.py +0 -0
  41. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/harness_shared/__init__.py +0 -0
  42. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/harness_shared/commands.py +0 -0
  43. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/harness_shared/integrity.py +0 -0
  44. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/harness_shared/paths.py +0 -0
  45. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/harness_shared/processes.py +0 -0
  46. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/hooks.json +0 -0
  47. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/mcp_config.json +0 -0
  48. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/pm/mcp_registration.jira.md +0 -0
  49. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/pm/mcp_registration.linear.md +0 -0
  50. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/rules/compose-context.md +0 -0
  51. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/rules/conflict-and-recovery.md +0 -0
  52. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/rules/gradle-context.md +0 -0
  53. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/rules/harness-rules.md +0 -0
  54. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/rules/kmp-context.md +0 -0
  55. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/rules/pm-and-context-bridge.md +0 -0
  56. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_adb_core.py +0 -0
  57. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_adb_core_selftest.py +0 -0
  58. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_android_inspect_selftest.py +0 -0
  59. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_android_scenarios_selftest.py +0 -0
  60. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_android_ui_selftest.py +0 -0
  61. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_antigravity_stability_selftest.py +0 -0
  62. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_architecture_selftest.py +0 -0
  63. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_baseline_selftest.py +0 -0
  64. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_change_decision_selftest.py +0 -0
  65. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_critical_safety_selftest.py +0 -0
  66. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_daily_workflow_selftest.py +0 -0
  67. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_debug_instrumentation_selftest.py +0 -0
  68. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_delivery_equivalence_selftest.py +0 -0
  69. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_efficiency_selftest.py +0 -0
  70. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_enforcement.py +0 -0
  71. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_env_codes.py +0 -0
  72. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_env_codes_selftest.py +0 -0
  73. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_environment.py +0 -0
  74. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_external_write_policy_selftest.py +0 -0
  75. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_gate_results.py +0 -0
  76. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_graph_core.py +0 -0
  77. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_graph_discovery_selftest.py +0 -0
  78. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_graph_selftest.py +0 -0
  79. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_hook_selftest.py +0 -0
  80. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_host_continuation_selftest.py +0 -0
  81. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_idempotency.py +0 -0
  82. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_installer_config.py +0 -0
  83. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_live_process.py +0 -0
  84. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_modules.py +0 -0
  85. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_performance_selftest.py +0 -0
  86. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_phase_review_v2_selftest.py +0 -0
  87. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_pm_selftest.py +0 -0
  88. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_portability_selftest.py +0 -0
  89. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_product.py +0 -0
  90. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_project_architecture.py +0 -0
  91. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_public_cli_selftest.py +0 -0
  92. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_public_commands.py +0 -0
  93. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_release_safety_selftest.py +0 -0
  94. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_repo_files.py +0 -0
  95. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_review_models_selftest.py +0 -0
  96. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_security_selftest.py +0 -0
  97. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_source_outline.py +0 -0
  98. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_source_text.py +0 -0
  99. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_stabilization_v41_selftest.py +0 -0
  100. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_stabilization_v42_selftest.py +0 -0
  101. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_task_isolation_selftest.py +0 -0
  102. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_variants.py +0 -0
  103. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_verification_recipes.py +0 -0
  104. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_verification_value_selftest.py +0 -0
  105. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_vnext_common.py +0 -0
  106. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_vnext_selftest.py +0 -0
  107. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_workflow/__init__.py +0 -0
  108. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_workflow/engine.py +0 -0
  109. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_workflow/phase_routing.py +0 -0
  110. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_workflow/router.py +0 -0
  111. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_workflow/verification.py +0 -0
  112. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_workflow_insight_selftest.py +0 -0
  113. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/_worktree_handoff_selftest.py +0 -0
  114. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/android_inspect.py +0 -0
  115. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/android_ui.py +0 -0
  116. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/antigravity_contract.py +0 -0
  117. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/antigravity_runtime.py +0 -0
  118. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/antigravity_tools.py +0 -0
  119. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/architecture_drift.py +0 -0
  120. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/architecture_policy.py +0 -0
  121. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/architecture_resolver.py +0 -0
  122. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/artifact_set.py +0 -0
  123. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/baseline_capture.py +0 -0
  124. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/build_execution.py +0 -0
  125. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/capture_screen.py +0 -0
  126. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/cc_pre_tool_safety.py +0 -0
  127. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/change_decision.py +0 -0
  128. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/check_kit_update.py +0 -0
  129. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/check_strings.py +0 -0
  130. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/context_presentation.py +0 -0
  131. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/copilot_pre_tool_safety.py +0 -0
  132. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/debug_device.py +0 -0
  133. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/debug_instrumentation.py +0 -0
  134. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/debug_source.py +0 -0
  135. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/delivery_equivalence.py +0 -0
  136. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/delivery_manifest.py +0 -0
  137. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/discovery_receipt.py +0 -0
  138. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/discovery_router.py +0 -0
  139. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/doctor/__init__.py +0 -0
  140. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/doctor/models.py +0 -0
  141. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/ensure_hook_selftest.py +0 -0
  142. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/evidence_store.py +0 -0
  143. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/exported_component_guard.py +0 -0
  144. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/fast_kt_lint.py +0 -0
  145. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/final_verdict.py +0 -0
  146. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/final_verifier.py +0 -0
  147. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/gradle_error_parser.py +0 -0
  148. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/harness_doctor.py +0 -0
  149. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/host_continuation.py +0 -0
  150. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/incident_reporter.py +0 -0
  151. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/install_or_update.py +0 -0
  152. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/install_tool_adapters.py +0 -0
  153. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/integrations/__init__.py +0 -0
  154. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/integrations/base.py +0 -0
  155. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/integrations/external_write_policy.py +0 -0
  156. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/integrations/generic_mcp.py +0 -0
  157. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/integrations/registry.py +0 -0
  158. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/integrations/universal_pm.py +0 -0
  159. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/integrations/zoho_sprints/__init__.py +0 -0
  160. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/integrations/zoho_sprints/integration.py +0 -0
  161. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/integrations/zoho_sprints/policy.py +0 -0
  162. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/journey_runner.py +0 -0
  163. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/lifecycle.py +0 -0
  164. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/logcat_doctor.py +0 -0
  165. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/maestro_bridge.py +0 -0
  166. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/mutation_guard.py +0 -0
  167. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/new_feature_scaffold.py +0 -0
  168. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/perf_guard.py +0 -0
  169. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/phone_e2e.py +0 -0
  170. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/plan_authority.py +0 -0
  171. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/plan_notes.py +0 -0
  172. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/pm_discovery.py +0 -0
  173. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/pm_github.py +0 -0
  174. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/pm_policy.py +0 -0
  175. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/policy_vocab.py +0 -0
  176. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/post_tool_reconcile.py +0 -0
  177. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/pre_tool_safety.py +0 -0
  178. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/preflight.py +0 -0
  179. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/preflight_check.py +0 -0
  180. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/record_review.py +0 -0
  181. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/repair.py +0 -0
  182. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/review_budget.py +0 -0
  183. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/review_execution.py +0 -0
  184. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/review_fix_delta.py +0 -0
  185. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/review_models.py +0 -0
  186. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/review_orchestrator.py +0 -0
  187. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/review_policy.py +0 -0
  188. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/review_reasoning.py +0 -0
  189. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/review_sources.py +0 -0
  190. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/reviewer_host.py +0 -0
  191. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/room_guard.py +0 -0
  192. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/run_device.py +0 -0
  193. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/run_gradle_task.py +0 -0
  194. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/run_tests_gate.py +0 -0
  195. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/self_heal.py +0 -0
  196. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/setup_wizard.py +0 -0
  197. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/skill_router.py +0 -0
  198. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/static_health_bridge.py +0 -0
  199. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/symbol_edit.py +0 -0
  200. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/task_base.py +0 -0
  201. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/task_bridge.py +0 -0
  202. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/task_git_lineage.py +0 -0
  203. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/task_merge.py +0 -0
  204. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/task_metrics.py +0 -0
  205. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/verification_contract.py +0 -0
  206. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/verification_steps.py +0 -0
  207. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/wizard/__init__.py +0 -0
  208. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/wizard/discovery.py +0 -0
  209. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/wizard/i18n.py +0 -0
  210. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/wizard/questions.py +0 -0
  211. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/wizard/schema.py +0 -0
  212. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/workflow.py +0 -0
  213. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/zoho_grant.py +0 -0
  214. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/scripts/zoho_sync.py +0 -0
  215. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/settings.json +0 -0
  216. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-harness/SKILL.md +0 -0
  217. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-harness/references/architecture-guidelines.md +0 -0
  218. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-harness/references/automated-skills.md +0 -0
  219. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-harness/references/command-contract.md +0 -0
  220. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-harness/references/daily-scenarios.md +0 -0
  221. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-harness/references/database-and-persistence.md +0 -0
  222. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-harness/references/performance-and-optimization.md +0 -0
  223. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-harness/references/test-quality-guidelines.md +0 -0
  224. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-harness/references/ui-layout-and-theming.md +0 -0
  225. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-security-billing/SKILL.md +0 -0
  226. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/android-ui-diagnostics/SKILL.md +0 -0
  227. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/brainstorming/SKILL.md +0 -0
  228. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/compose-inspector/SKILL.md +0 -0
  229. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/efficient-context/SKILL.md +0 -0
  230. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/git-pr-automator/SKILL.md +0 -0
  231. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/gradle-build-optimizer/SKILL.md +0 -0
  232. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/kotlin-coroutines-expert/SKILL.md +0 -0
  233. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/phone-e2e/SKILL.md +0 -0
  234. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/pm-tracker/SKILL.md +0 -0
  235. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/systematic-debugging/SKILL.md +0 -0
  236. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/skills/test-driven-development/SKILL.md +0 -0
  237. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/subagents/android-ui-expert-agent.json +0 -0
  238. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/subagents/arch-regression-reviewer-agent.json +0 -0
  239. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/subagents/bug-test-reviewer-agent.json +0 -0
  240. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/subagents/qa-diagnostics-agent.json +0 -0
  241. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/subagents/re-review-prompt.md +0 -0
  242. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/subagents/review-prompt.md +0 -0
  243. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/subagents/security-perf-reviewer-agent.json +0 -0
  244. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/subagents/spec-compliance-agent.json +0 -0
  245. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/AGENTS.md.template +0 -0
  246. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/CLAUDE.md.template +0 -0
  247. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/CODEX.md.template +0 -0
  248. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/GEMINI.md.template +0 -0
  249. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/QWEN.md.template +0 -0
  250. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/README.md +0 -0
  251. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/continue-android-harness.md.template +0 -0
  252. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/copilot-instructions.md.template +0 -0
  253. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/cursor-android-harness.mdc.template +0 -0
  254. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/gemini-runtime/android-harness-global.md.template +0 -0
  255. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/gemini-runtime/config.grants.example.json +0 -0
  256. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/gemini-runtime/mcp_config.template.json +0 -0
  257. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/github-instructions.md.template +0 -0
  258. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/pointer.md.template +0 -0
  259. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/tool-adapters/windsurf-android-harness.md.template +0 -0
  260. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/android-inspection.md +0 -0
  261. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/android-ui.md +0 -0
  262. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/check-strings.md +0 -0
  263. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/commit-msg.md +0 -0
  264. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/crash-triage.md +0 -0
  265. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/debug.md +0 -0
  266. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/deliver.md +0 -0
  267. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/efficient-context.md +0 -0
  268. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/new-feature.md +0 -0
  269. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/perf-audit.md +0 -0
  270. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/phone-e2e.md +0 -0
  271. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/pm-tracker.md +0 -0
  272. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/preflight.md +0 -0
  273. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/agents/workflows/test-quality-audit.md +0 -0
  274. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/android_agent_harness.egg-info/SOURCES.txt +0 -0
  275. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/android_agent_harness.egg-info/dependency_links.txt +0 -0
  276. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/android_agent_harness.egg-info/entry_points.txt +0 -0
  277. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/android_agent_harness.egg-info/top_level.txt +0 -0
  278. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/harness_cli.py +0 -0
  279. {android_agent_harness-1.3.2 → android_agent_harness-1.3.4}/setup.cfg +0 -0
@@ -0,0 +1,455 @@
1
+ Metadata-Version: 2.4
2
+ Name: android-agent-harness
3
+ Version: 1.3.4
4
+ Summary: Approval-first, evidence-bound engineering harness for Android and KMM/KMP (Kotlin Multiplatform Mobile) projects across Google Antigravity, Claude Code, Codex, Cursor, and AI coding agents.
5
+ Author: Rabee Elkholy
6
+ License-Expression: MIT
7
+ Project-URL: Homepage, https://github.com/rabee-elkholy/android-agent-harness
8
+ Project-URL: Issues, https://github.com/rabee-elkholy/android-agent-harness/issues
9
+ Project-URL: Source, https://github.com/rabee-elkholy/android-agent-harness
10
+ Project-URL: Documentation, https://github.com/rabee-elkholy/android-agent-harness/tree/main/docs
11
+ Project-URL: Changelog, https://github.com/rabee-elkholy/android-agent-harness/blob/main/CHANGELOG.md
12
+ Project-URL: Security, https://github.com/rabee-elkholy/android-agent-harness/blob/main/SECURITY.md
13
+ Keywords: android,kmm,kmp,kotlin-multiplatform,antigravity,claude-code,ai-agents,code-review,governance,mcp
14
+ Classifier: Development Status :: 5 - Production/Stable
15
+ Classifier: Environment :: Console
16
+ Classifier: Intended Audience :: Developers
17
+ Classifier: Operating System :: OS Independent
18
+ Classifier: Programming Language :: Python :: 3
19
+ Classifier: Programming Language :: Python :: 3.10
20
+ Classifier: Programming Language :: Python :: 3.11
21
+ Classifier: Programming Language :: Python :: 3.12
22
+ Classifier: Programming Language :: Python :: 3.13
23
+ Classifier: Programming Language :: Python :: 3.14
24
+ Classifier: Topic :: Software Development :: Quality Assurance
25
+ Requires-Python: >=3.10
26
+ Description-Content-Type: text/markdown
27
+ License-File: LICENSE
28
+ Dynamic: license-file
29
+
30
+ <div align="center">
31
+
32
+ <img src="https://raw.githubusercontent.com/rabee-elkholy/android-agent-harness/main/docs/assets/banner.svg" alt="Android Agent Harness — Stricter Proof, Not Heavier Workflow" width="100%" />
33
+
34
+ <br />
35
+
36
+ [![CI](https://github.com/rabee-elkholy/android-agent-harness/actions/workflows/ci.yml/badge.svg)](https://github.com/rabee-elkholy/android-agent-harness/actions/workflows/ci.yml)
37
+ [![Release](https://img.shields.io/github/v/release/rabee-elkholy/android-agent-harness?label=release&color=3DDC84)](https://github.com/rabee-elkholy/android-agent-harness/releases)
38
+ [![PyPI](https://img.shields.io/pypi/v/android-agent-harness?color=38BDF8)](https://pypi.org/project/android-agent-harness/)
39
+ [![Targets](https://img.shields.io/badge/Android%20%7C%20KMM%20%2F%20KMP-Kotlin%20%7C%20Java%20%7C%20Compose%20%7C%20Swift-7F52FF)](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/compatibility-matrix.md)
40
+ [![Python](https://img.shields.io/badge/Python-3.10%E2%80%933.14%20%7C%20Zero%20Deps-2563EB)](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/compatibility-matrix.md)
41
+ [![Platforms](https://img.shields.io/badge/Linux%20%7C%20macOS%20%7C%20Windows-supported-10B981)](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/compatibility-matrix.md)
42
+ [![License: MIT](https://img.shields.io/github/license/rabee-elkholy/android-agent-harness)](https://github.com/rabee-elkholy/android-agent-harness/blob/main/LICENSE)
43
+ [![XERJ Corpus Hub contributor](https://hub.xerj.org/contributors/rabee-elkholy/badge.svg)](https://hub.xerj.org/contributors/rabee-elkholy)
44
+
45
+ <h3>An approval-first, cryptographically verified engineering layer for AI coding agents on Native Android &amp; Kotlin Multiplatform (KMM / KMP).</h3>
46
+
47
+ <p>
48
+ <a href="#quick-start"><b>Quick Start</b></a> ·
49
+ <a href="#why-android-agent-harness"><b>Why Harness?</b></a> ·
50
+ <a href="#what-daily-use-looks-like"><b>Daily Workflow</b></a> ·
51
+ <a href="#antigravity-first-architecture--review-protocol-v2"><b>Antigravity &amp; Review V2</b></a> ·
52
+ <a href="#hybrid-flash-orchestration--frontier-quota-protection"><b>Hybrid Models &amp; Quota</b></a> ·
53
+ <a href="#capability-detected-android-ecosystem-bridges--ast-symbol-surgery"><b>Ecosystem Bridges</b></a> ·
54
+ <a href="#measured-performance--context-efficiency"><b>Benchmarks</b></a> ·
55
+ <a href="#faq"><b>FAQ</b></a>
56
+ </p>
57
+
58
+ </div>
59
+
60
+ ---
61
+
62
+ **Android Agent Harness** turns general-purpose AI coding assistants — primary-first on **Google Antigravity**, plus **Claude Code**, **Codex**, **Cursor**, and `AGENTS.md` hosts — into a disciplined, senior-grade Android and KMM/KMP engineering workflow.
63
+
64
+ Instead of relying on fragile prompt instructions that decay as chat context grows, the harness moves architectural discovery, scope boundaries, TDD bug reproduction, specialist code review, physical-device QA, and delivery proof into a **zero-dependency Python standard-library runtime** living cleanly under `.agents/`.
65
+
66
+ > **Core Philosophy — Stricter proof, not heavier daily workflow:** You approve one clear plan, keep 100% authority over Git and external mutations, and receive changes backed by SHA-256 snapshot-bound compile, test, review, and device evidence.
67
+
68
+ ---
69
+
70
+ ## At a Glance: 10 Core Engineering Capabilities
71
+
72
+ | Capability | Problem Solved | How Android Agent Harness Enforces It |
73
+ |---|---|---|
74
+ | **1. AST & Symbol Graph Discovery** | Blind grep loops and token-heavy full-file reads across large multi-module repos | Incremental code graph (`graph` & `task-context`) across `Kotlin`, `Java`, `XML`, and `Swift`, surfacing exact line ranges, callers, reuse candidates, and matching tests in `~14 ms`. |
75
+ | **2. Single-Approval Scope Guard** | Silent scope creep, unapproved refactors, and repetitive confirmation prompts | One plan approval (`plan_sha256` + single-use nonce) unlocks execution. Native `PreToolUse` hooks (`hooks.json`) block unapproved writes in real time. |
76
+ | **3. Multi-Wave Hybrid Flash Delegation** | Burning weekly frontier model quota on file exploration, boilerplate slices, and Gradle loops | Pinned `gemini-3.8-flash-medium` `harness-scout-agent` and `harness-worker-agent` absorb discovery and `T0–T3` slices with hard step caps (`24`/`32`) and Lead anti-reread guards (`<=3` reads). |
77
+ | **4. 6-Tier Risk & Epistemic Review V2** | Over-reviewing trivial docs or under-reviewing `Auth` / `Billing` / `Room` changes | Classifies actual diffs from `T0_TRIVIAL` (instant delivery) to `T5_CRITICAL`, dispatching epistemically isolated specialists returning machine-verified `HARNESS_REVIEW_RESULT_V2` verdicts. |
78
+ | **5. Android & KMM / KMP Domain Gates** | Room migration crashes, missing `values-*` translations, layer bypasses, and `commonMain` leaks | Deterministic preflight checks for Room schema bumps & migrations, locale string/plural parity, Clean Architecture `UI -> DAO` bypass, Android 15/16 lint, and `commonMain` purity. |
79
+ | **6. Persistent Repo Memory (`.agents/project-context/`)** | Repeating team conventions and architectural rules in every new chat | Combines machine-extracted `project-facts.json`, human `project-notes.md`, and scoped `developer-instructions.json` propagated across all `7/7` discovery, planning, worker, and review surfaces. |
80
+ | **7. AST Symbol Surgery & Ecosystem Bridges** | Fragile 600-line file edits, RTL clipping missed by unit tests, and noisy repo-wide lint | Built-in declaration-level AST editor (`symbol replace-body`), **Maestro YAML** physical-device runner & exporter, **Roborazzi** headless `.uitree.json` RTL/bounds gate, and diff-scoped **SARIF / BuildHealth / LeakCanary** checks. |
81
+ | **8. Cryptographic Snapshot & Device Proof** | Claiming "tests pass" from stale runs before the latest edit | Binds unit tests, reviews, APK artifacts, `device install-start`, and live phone E2E journeys to `delivery_snapshot_sha256` and `change_set_sha256`. Any edit invalidates stale proof. |
82
+ | **9. Smart Context Rotation & Task Bridge** | High latency in `100k+`-token chats and re-explaining context across follow-up tasks | Tracks session health, recommends clean rotation boundaries, and transfers contracts, `interface_delta` signatures, and a fresh Discovery Receipt to a new chat via `task bridge`. |
83
+ | **10. Universal PM MCP & Self-Healing State** | Accidental ticket closures, overwritten Bug descriptions, or wedged task states | Zero-config discovery for Zoho Sprints/Projects, Jira, Linear, GitHub, ClickUp, Asana, Monday, and Trello with `pm authorize` previews, protected Bug bodies, `task heal` rollback, and Zero-PII forensics. |
84
+
85
+ ---
86
+
87
+ ## Why Android Agent Harness?
88
+
89
+ AI models write Kotlin and Compose code fast — keeping them aligned with a real, evolving multi-module Android codebase over dozens of turns is the hard part.
90
+
91
+ | Failure Mode in Vanilla AI Sessions | Deterministic Guarantee with Android Agent Harness |
92
+ |---|---|
93
+ | **Stale verification**: Agent reports "all tests pass" from a run before its last edit. | Every test, review, APK, and device check is bound to `delivery_snapshot_sha256`. Any later edit marks proof stale and blocks delivery. |
94
+ | **Unreproduced bug fixes**: Agent guesses at a fix without proving the bug existed. | Applicable `BUG` tasks enforce `CAPTURE_RED_EVIDENCE` (`--capture-red`) so a failing test reproduces the defect before production code is touched. |
95
+ | **Scope drift & drive-by refactors**: Agent edits unrelated modules or modernizes legacy Java/XML unasked. | Local architecture contracts preserve per-module patterns (`MVP`, `MVVM`, `MVI`), and native `PreToolUse` hooks block out-of-scope file mutations. |
96
+ | **Self-approved plans & unwanted Git commits**: Agent rubber-stamps its own plan or runs `git commit`/`push`. | Plan approval requires developer words in chat; `commit`, `branch`, `merge`, `rebase`, and `push` remain strictly under **Human Git Authority**. |
97
+ | **Frontier quota exhaustion**: Lead model re-reads 800-line files after subagents finish or spawns unpinned explore loops. | Built-in unpinned `research`/`explore` loops are redirected to Flash `harness-scout-agent`, and post-wave Lead re-reads are capped at `<=3` surgical checks. |
98
+ | **Context amnesia across chats**: Every new conversation forgets module rules and past architectural decisions. | `project-notes.md` and scoped `developer-instructions.json` live in the repo, automatically rehydrating constraints and checking code-to-doc drift (`context drift-report`). |
99
+
100
+ ---
101
+
102
+ ## Quick Start
103
+
104
+ ### Chat installation (recommended)
105
+
106
+ Open your Android or KMM/KMP project in **Google Antigravity** (or Claude Code) and paste this single bootstrap prompt:
107
+
108
+ ```text
109
+ Read https://raw.githubusercontent.com/rabee-elkholy/android-agent-harness/v1.3.4/docs/install-or-update-prompt.md and follow all instructions.
110
+ ```
111
+
112
+ What happens automatically and non-destructively:
113
+ 1. **Read-only inspection**: Discovers Gradle modules (`:app`, `:composeApp`, `:core:*`, `:feature:*`), flavors, UI toolkits, and locales without modifying a single app file.
114
+ 2. **Setup preview & approval**: Asks essential project questions, previews the installation plan, and waits for your approval.
115
+ 3. **Clean isolation**: Stores the pinned kit outside your repo (`~/.android-harness/kit`), installs only the managed runtime under `.agents/`, configures `.git/info/exclude` (zero `git status` noise when gitignored mode is chosen), and verifies installation health via `doctor --install-check`.
116
+
117
+ ### Terminal installation (alternative)
118
+
119
+ Install globally via `pipx`:
120
+
121
+ ```bash
122
+ # 1. Install the standalone CLI
123
+ pipx install android-agent-harness
124
+
125
+ # 2. Initialize inside your Android / KMP repository
126
+ android-harness init --repo /path/to/android-project
127
+
128
+ # 3. Run the 13-dimension health diagnostic
129
+ android-harness doctor --repo /path/to/android-project --json
130
+ ```
131
+
132
+ Or run directly from a pinned shallow clone with **zero global packages**:
133
+
134
+ ```bash
135
+ git clone --depth 1 --branch v1.3.4 --single-branch \
136
+ https://github.com/rabee-elkholy/android-agent-harness.git ~/.android-harness/kit
137
+
138
+ python ~/.android-harness/kit/harness_cli.py init \
139
+ --repo /path/to/android-project \
140
+ --kit ~/.android-harness/kit
141
+ ```
142
+
143
+ Inside an installed repository, all daily operations run through the in-process local launcher:
144
+
145
+ ```bash
146
+ python .agents/harness.py doctor --json
147
+ ```
148
+
149
+ ### Transactional Zero-Clobber Updates
150
+
151
+ Installed repositories check for newer releases in the background (at most once per day) and provide the exact upgrade command via `python .agents/harness.py update-info`:
152
+
153
+ ```bash
154
+ # Update from a pinned tag checkout:
155
+ python ~/.android-harness/kit/harness_cli.py update --repo /path/to/android-project
156
+
157
+ # Or update from a local maintainer working branch:
158
+ python /path/to/android-agent-harness/harness_cli.py update --repo /path/to/android-project --kit /path/to/android-agent-harness --no-refresh
159
+ ```
160
+
161
+ Updates synchronize `.agents/` in place (friendly to open IDE file watchers), preserve `project-notes.md`, `developer-instructions.json`, setup answers, and task history, refuse by default while a task is mid-flight (unless `--preserve-active-task <id> --source-stopped` is explicitly requested), and automatically roll back if interrupted. See [Install & Update Guide](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/install-or-update-prompt.md).
162
+
163
+ ---
164
+
165
+ ## What daily use looks like
166
+
167
+ <p align="center">
168
+ <img src="https://raw.githubusercontent.com/rabee-elkholy/android-agent-harness/main/docs/assets/architecture-pipeline.svg" alt="Android Agent Harness 5-Stage Pipeline" width="100%" />
169
+ </p>
170
+
171
+ You write natural engineering requests; the harness keeps state, routing, and verification deterministic behind the scenes:
172
+
173
+ > **Developer:** *"Fix the calories calculation bug in `CaloriesViewModel` when `weightGoal` is blank for `STEPS` users."*
174
+
175
+ 1. **Anchored Discovery**: Resolves exact callers, tests, and repo-local constraints via `graph` and `task-context` (or parallel Flash `harness-scout-agent` waves on multi-subsystem tasks).
176
+ 2. **One Concise Plan**: Generates a readable `plan.md` and presents a language-adaptive approval card via `ask_question`.
177
+ 3. **Explicit Developer Approval**: Waits once for your approval (`Approve` / `Request changes` / `Cancel task`) — no mid-task nagging.
178
+ 4. **Scoped Execution**: Follows the deterministic router (`task status --task-id <id> --next`), reproduces the bug (`--capture-red`), and applies surgical edits strictly inside approved files.
179
+ 5. **Risk-Adaptive Gates**: Runs preflight checks, targeted module unit tests, and epistemically isolated V2 specialist reviews matched to the final diff's risk tier.
180
+ 6. **Device & Walkthrough Sign-off**: Installs/launches on a connected device when required (`device install-start` or `maestro run`) and presents a concise mobile verification card.
181
+ 7. **Human Git Delivery**: Leaves verified changes unstaged (or ready for your commit) and closes out cleanly once you commit.
182
+
183
+ ### Multi-Phase Tasks, Disjoint Parallel Tasks & Smart Context Bridging
184
+
185
+ - **Coherent Multi-Phase Execution**: Large features execute across semantic phases (`p1 -> p2 -> ...`) under **one initial approval**. Phase boundaries run fast deterministic checkpoints (`preflight`, Room guards, string parity, fast ktlint, and targeted unit tests) without pausing for mid-phase AI subagents unless `scoped_phase_review_enabled` is explicitly enabled, followed by a full final integration review.
186
+ - **Up to 3 Live Disjoint Tasks per Worktree**: Keep up to three independent tasks open simultaneously on non-overlapping files (`task switch --task-id <id>`), combine pre-approval plans safely (`task merge`), or freeze urgent hotfix interruptions with cryptographic lineage (`task handoff` -> developer WIP commit -> `task reconcile-handoff`).
187
+ - **Smart Context Rotation & `task bridge`**: Long chats (`>=120` steps / `>=80k` tokens at task boundaries) surface a non-blocking rotation advisory. One command transfers full architectural context, `interface_delta` signatures, and a fresh conversation-local Discovery Receipt to a clean chat — or across AI hosts (`task continue --host <host> --source-stopped`):
188
+
189
+ ```bash
190
+ # Save follow-up or mid-task handoff state in the current chat:
191
+ python .agents/harness.py task bridge --save \
192
+ --next-outcome "Wire CaloriesSettingsScreen state into HomeMviScreen" \
193
+ --summary "Completed CaloriesViewModel maintenance equation; preserve UiState contract"
194
+
195
+ # Restore immediately in a fresh chat with zero re-discovery overhead:
196
+ python .agents/harness.py task bridge
197
+ ```
198
+
199
+ ---
200
+
201
+ ## Antigravity-First Architecture & Review Protocol V2
202
+
203
+ Android Agent Harness is engineered primary-first for **Google Antigravity** with `HARD_ENFORCED` native hooks (`.agents/hooks.json`), custom subagent definitions under `.agents/agents/<role>/agent.md`, and native transcript execution verification:
204
+
205
+ - **Machine-Verified `HARNESS_REVIEW_RESULT_V2`**: Reviewers return structured JSON payloads cryptographically bound to `delivery_snapshot_sha256`, `change_set_sha256`, and `package_sha256` — eliminating prose-parsing hallucinations.
206
+ - **Epistemically Isolated Specialist Roster (`.agents/agents/`)**:
207
+ - `bug-test-reviewer-agent` — Behavioral logic, edge cases, nullability, coroutine cancellation/structured concurrency, lifecycle state, and unit test assertion strength.
208
+ - `arch-regression-reviewer-agent` — Caller/navigation/schema blast radius, Clean Architecture layer boundaries, KMP source-set purity, localization parity, and codebase reuse.
209
+ - `security-perf-reviewer-agent` — Auth/Billing/Crypto safety, exported components, `PendingIntent` mutability, main-thread I/O & ANR hazards, Compose stability, and FGS/sensor leaks.
210
+ - `spec-compliance-agent` — Requirement-by-requirement verification against the approved plan when routed.
211
+ - **`Zero-Hop` Inlined Briefs & Single-Turn Batch Completion**: Canonical review packages (`<= 24 KB` / `520` lines), pre-computed `GraphEngine` blast-radius topologies, and already-passed deterministic gate summaries (`preflight`, `compile`, `unit_tests`) are embedded directly inside `brief-<role>.md` (and `phase-brief-<role>.md`). Once parallel reviewers finish, `python .agents/harness.py review complete-all --task <id>` ingests all verdicts in a single turn.
212
+ - **Flexible Reviewer Models & Reviewer Call Safety Cap**: Supports default parent inheritance, interactive presets (`risk-hybrid`, `flash-dynamic`, `carbon-dynamic`), or explicit Antigravity/Claude model pins, guarded by a configurable **Reviewer Call Safety Cap** (default: `20`, per-reviewer tool budget: `12` calls). See [Reviewer Model Configuration](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/reviewer-models.md).
213
+
214
+ ---
215
+
216
+ ## Hybrid Flash Orchestration & Frontier Quota Protection
217
+
218
+ Large Android repositories can drain weekly frontier quotas (`Gemini 4 Carbon` / `Argon` / `Claude 5.5`) if the primary model performs every multi-file search, boilerplate edit, and Gradle test loop itself. Version `1.3.2` enforces **project-wide multi-wave subagent governance**:
219
+
220
+ | Layer / Role | Pinned / Recommended Model | Responsibility & Enforced Guardrails |
221
+ |---|---|---|
222
+ | **Lead Architect** (Main Chat) | `Gemini 4 Carbon (Medium)` / `Argon` ⇄ rotate to `Claude Sonnet/Opus 5.5` at `~50%` weekly bar | Architectural reasoning, interactive `ask_question` alignment, `T4/T5` sensitive production code (`AUTH`, `BILLING`, `SECURITY`, `CRYPTO`), semantic conflict resolution, and final `git diff` inspection. |
223
+ | **`harness-scout-agent`** (`1–3` parallel) | `gemini-3.8-flash-medium` | Read-only `graph` / `task-context` anchoring and parallel `view_file` sweeps returning `<1,500`-token `file:line` digests. Hard hook cap: **`24` tool calls/instance**, max **`2` waves/turn**. Unpinned built-in `research`/`explore` subagents are blocked (`USE_HARNESS_SCOUT_AGENT`). |
224
+ | **`harness-worker-agent`** (Single active) | `gemini-3.8-flash-medium` | Executes decision-complete slices via `delegate-plan --json` (`FULL_SLICE` on `T0–T3`, `CAPTURE_RED_ONLY` on `BUG` pre-RED, `TESTS_AND_GATES_ONLY` on `T4/T5` & `CONFLICT`). Hard hook cap: **`32` read/search calls**; edits and `compile`/`test` gates uncapped. |
225
+ | **Post-Wave Anti-Reread Guard** | Native `PreToolUse` + `PostToolUse` | Blocks bloated `>220`-line reads on `>350`-line files after subagent waves and caps Lead post-wave checks at **`<=3` surgical reads** (`0` redundant re-reads after `APPROVE` reviews). |
226
+
227
+ ---
228
+
229
+ ## Adaptive 6-Tier Risk Model & Android / KMP Protection
230
+
231
+ Verification scales strictly with the **classified final diff** — never slowing down documentation or string tweaks while never cutting corners on database migrations or billing flows:
232
+
233
+ | Tier | Classification | Representative Changes | Routed Verification Profile |
234
+ |---|---|---|---|
235
+ | **T0** | `TRIVIAL` | Markdown docs, comments, non-runtime prose | Fast structural / preflight check — zero reviewers, instant delivery |
236
+ | **T1** | `LOW_RISK` | `strings.xml`, colors/dimens, static resource tweaks | Localized string/plural parity + bounded compile check as needed |
237
+ | **T2** | `FEATURE` | `ViewModel`, `UseCase`, structured `Flow`/`suspend`, Compose/XML UI, `DataStore` | Module unit tests + targeted V2 review (consolidated single reviewer on isolated `<=25`-line private helper edits) + APK assemble |
238
+ | **T3** | `SUBSYSTEM` | Multi-module contracts, Hilt/Koin DI, Retrofit/Ktor networking, custom dispatchers | Integration & architecture boundary review + cross-module compile/test gates |
239
+ | **T4** | `DATA_DEVICE` | Room / SQLDelight schemas & migrations, `ForegroundService`, Camera, Bluetooth, Sensors | Room migration guard + specialist reviews + physical device / Maestro proof |
240
+ | **T5** | `CRITICAL` | `Auth`, `Billing`, `Crypto`, `Keystore`, permissions, sensitive user data | Full specialist roster (`risk-hybrid` escalation eligible) + explicit sensitive approval |
241
+
242
+ ### Native Android & Kotlin Multiplatform (`KMM / KMP`) Safeguards
243
+ - **Multiplatform Source-Set Purity**: Indexes `commonMain`, `androidMain`, `iosMain`, `:composeApp`, `:shared`, and iOS `.swift` symbols; blocks accidental `android.*` / JVM-only API leaks in `commonMain`; and validates `expect` / `actual` declaration pairing.
244
+ - **Mixed Legacy & Modern Codebases**: Resolves local architecture profiles per directory/module (`Java + XML + MVP` alongside `Kotlin + XML + MVVM` and `Kotlin + Compose + MVI`) without forcing unapproved rewrites.
245
+ - **Android 15 / 16 & Platform Readiness**: Zero-token static checks in `fast_kt_lint.py` catch missing `PendingIntent` immutability flags, swallowed `CancellationException` in `suspend` functions, hardcoded UI strings, UI-to-DAO layer bypasses, and exported Manifest components missing permissions.
246
+
247
+ ---
248
+
249
+ ## Persistent Project Knowledge (`.agents/project-context/`)
250
+
251
+ > **Your repository remembers architectural truth across every chat, model, and teammate.**
252
+
253
+ ```text
254
+ 1. Current Source Code Evidence (Highest authority — live AST & working tree)
255
+ ↓
256
+ 2. Generated Project Facts (.agents/project-context/project-facts.json + architecture/ui/persistence/conventions.md)
257
+ ↓
258
+ 3. Human Notes & Instructions (project-notes.md + scoped developer-instructions.json across all 7/7 surfaces)
259
+ ↓
260
+ 4. Generic Android Guidance (Lowest authority — never overrides local codebase evidence)
261
+ ```
262
+
263
+ ```bash
264
+ # Record an idempotent team convention in .agents/project-context/project-notes.md:
265
+ python .agents/harness.py context note \
266
+ --section "Domain Conventions & Context" \
267
+ "Never subtract 500 kcal when weightGoal is empty on STEPS users; default to maintenance calories."
268
+
269
+ # Record a cryptographically bound, scope-matched instruction in developer-instructions.json:
270
+ python .agents/harness.py context instruct \
271
+ "Preserve the callback bridge in payments; do not migrate to Flow unless explicitly requested." \
272
+ --scope "MODULE::payments" \
273
+ --strength REQUIREMENT \
274
+ --source conversation \
275
+ --proof-reference "Keep payments callback bridge"
276
+
277
+ # Audit project-notes.md, developer-instructions.json, and project-facts.json for stale paths/modules:
278
+ python .agents/harness.py context drift-report --json
279
+ ```
280
+
281
+ Supported instruction scopes: `GLOBAL`, `MODULE`, `SOURCE_SET`, `PACKAGE`, `FEATURE`, `ARCH_FAMILY`, and `PATH`. Active instructions propagate automatically across `PreInvocation` reminders, `graph`, `task-context`, `plan.md`, `delegate-plan`, `phase_review`, and final `review_package` briefs, while non-code PM/Zoho HTML templates in `project-notes.md` are automatically stripped from code discovery and code-review prompts (`-97.8%` note payload bloat).
282
+
283
+ ---
284
+
285
+ ## Capability-Detected Android Ecosystem Bridges & AST Symbol Surgery
286
+
287
+ Zero external dependencies are forced on your project: tools that aren't configured add **`0 ms`** overhead, while detected capabilities unlock snapshot-bound verification automatically.
288
+
289
+ ### 1. Pure-Stdlib AST Symbol Surgery (`Kotlin`, `Java`, `Swift`)
290
+ Inspect and edit individual declarations inside 800-line classes without loading surrounding boilerplate — complete with multiline raw-string/comment scrubbing, `--expected-sha256` optimistic concurrency, whole-file brace-balance verification, and `PreToolUse` scope enforcement:
291
+
292
+ ```bash
293
+ python .agents/harness.py symbol outline --file app/src/main/java/.../ProfileViewModel.kt
294
+ python .agents/harness.py symbol read --file app/src/main/java/.../ProfileViewModel.kt --symbol ProfileViewModel.refresh
295
+ python .agents/harness.py symbol replace-body \
296
+ --file app/src/main/java/.../ProfileViewModel.kt \
297
+ --symbol ProfileViewModel.refresh \
298
+ --expected-sha256 <symbol_sha256> \
299
+ --replacement-file <path>
300
+ ```
301
+
302
+ ### 2. Maestro YAML Physical-Device Bridge & Live E2E Exporter (`maestro`)
303
+ - Runs `.maestro/*.yaml` flows against the bound physical device (`python .agents/harness.py maestro run --flow .maestro/login.yaml`) and binds results to `delivery_snapshot_sha256`.
304
+ - Blocks destructive YAML commands (`clearState`, `clearKeychain`, `uninstallApp`) fail-closed unless explicitly authorized (`--allow-clear-state`).
305
+ - Compiles passing interactive `phone_e2e` sessions into repeatable `.maestro/<task_id>.yaml` regression flows (`maestro export-e2e`).
306
+
307
+ ### 3. Roborazzi Headless `.uitree.json` Viewport & RTL Bounds Gate (`roborazzi`)
308
+ - Auto-detects `io.github.takahirom.roborazzi` (including Gradle Version Catalog `libs.plugins.*` aliases) and audits `.uitree.json` sidecars on the JVM for `HORIZONTAL_VIEWPORT_CLIP` (catching Arabic RTL & locale clipping), `ZERO_DIMENSION_NODE`, and `CLICKABLE_SIBLING_COLLISION`, injecting verified UI bounds digests directly into reviewer briefs.
309
+
310
+ ### 4. Diff-Scoped Static Health (`SARIF` / `BuildHealth` / `LeakCanary`) & Session Forensics
311
+ - **`health report`**: Filters Detekt / Android Lint `*.sarif` and Gradle `build-health-report.json` findings strictly to task-modified files and modules, plus post-E2E `LeakCanary` heap-leak checks on debug builds.
312
+ - **`diagnose-session`**: Audits session transcripts and `.agents/state/audit_log.jsonl` for hook-denial streaks, read sprawl, and peak token steps with Zero-PII bundle export (`--export-bundle`).
313
+
314
+ ### 5. Universal Zero-Config Multi-Provider PM MCP & Self-Healing (`pm` & `task heal`)
315
+ - Auto-discovers installed MCP servers for **Zoho Sprints**, **Zoho Projects**, **Jira**, **Linear**, **GitHub Projects**, **ClickUp**, **Asana**, **Monday.com**, and **Trello**.
316
+ - Enforces preview-and-approval (`pm authorize`), protects original Bug descriptions from overwrite, blocks unauthorized `Done`/`Closed` transitions, and generates 5-section RTL/LTR QA handoff reports (`pm prepare-report`) with zero code identifier leaks.
317
+ - Captures last-known-good state snapshots at every lifecycle boundary for one-command recovery (`task heal --prepare-fresh-run`) and consent-first Zero-PII incident diagnostics (`report-issue`).
318
+
319
+ ---
320
+
321
+ ## Measured Performance & Context Efficiency
322
+
323
+ All runtime operations execute in pure Python 3.10–3.14 standard library with content-addressed `blake2b` caching and in-process CLI dispatch:
324
+
325
+ | Benchmark / Measurement | Baseline (`v1.3.1` / Unoptimized) | Current (`v1.3.2`) | Measured Gain |
326
+ |---|---:|---:|---:|
327
+ | Read-only `harness.py` CLI dispatch (in-process fast path) | `145.1 ms` | **`18.1 ms`** | **8.02x faster** |
328
+ | Warm `workflow.py revise` on 19-module / 760-file repo | `564.0 ms` | **`109.4 ms`** | **5.15x faster** |
329
+ | `task-context` XML/Gradle `FILE_FALLBACK` JSON payload | `40,394 B` | **`4,694 B`** | **-88.38% bytes** |
330
+ | `harness.py graph --module :app --format json` payload | `16,689 B` | **`2,130 B`** | **-87.24% bytes** |
331
+ | Non-PM code reviewer `project-notes.md` block (with Zoho HTML) | `15,408 B` | **`341 B`** | **-97.79% bytes** |
332
+ | Repeated `task-context --since <digest>` (`UNCHANGED` mode) | `22,568 B` | **`11,233 B`** | **-50.23% bytes** |
333
+ | Warm `resolve_task_contexts` batch (`mixed` / `kmp` / `+500` files) | `14.59 ms` | **`14.14 ms` / `14.46 ms` / `47.39 ms`** | **`1` facts read/batch** |
334
+ | Warm `resolve_next_action` router (`1` doc / `500` docs) | `8.77 ms` / `14.64 ms` | **`8.74 ms` / `14.46 ms`** | **Sub-15 ms deterministic** |
335
+ | Always-loaded instruction budget (`harness-rules` + `AGENTS` + `GEMINI`) | `36,000` char ceiling | **`24,372` chars (`67.7%`)** | **`11,628` chars headroom** |
336
+
337
+ Full methodology, reproducible scripts, and historical client certification reports (`Automattic/pocket-casts-android` frozen runs) are published in:
338
+ - [v1.3.2 Engine, Project-Context & Subagent Benchmarks](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/benchmark/v1.3.2-engine-and-context-benchmarks.md)
339
+ - [Context Summary Compaction Measurements (JSON)](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/implementation/context-summary-measurements.json)
340
+ - [End-to-End Client Benchmark Status](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/benchmark/benchmark-status.md)
341
+
342
+ ---
343
+
344
+ ## Supported Hosts & Enforcement Model
345
+
346
+ | AI Coding Host | Installed Adapters & Assets | Scope & Mutation Enforcement | Review Protocol | Final Snapshot Verifier |
347
+ |---|---|---|---|---|
348
+ | **Google Antigravity** *(Primary Target)* | `GEMINI.md`, `AGENTS.md`, `.agents/hooks.json`, `.agents/agents/*/agent.md` (4 V2 reviewers + 2 Flash utility subagents), rules, skills | **`HARD_ENFORCED`** (`PreToolUse` & `PostToolUse` native hooks) | Trusted transcript `HARNESS_REVIEW_RESULT_V2` | **Yes** (SHA-256 bound) |
349
+ | **Claude Code** | `CLAUDE.md`, `AGENTS.md`, `.claude/settings.json` hooks, `.claude/agents/` subagents | Hook + Rule enforced | Native subagent transcript verification | **Yes** (SHA-256 bound) |
350
+ | **Cursor / Codex / Windsurf / Copilot / `AGENTS.md` Hosts** | `AGENTS.md` + host rule adapters (`.cursor/rules`, `CODEX.md`, `.windsurf/rules`, `.github/copilot-instructions.md`) | **`RULE_ENFORCED`** + downstream CLI gates | Structured response ingestion | **Yes** (SHA-256 bound) |
351
+
352
+ See [AI Tool Support Matrix](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/tool-support.md) and [Cross-Host Continuation](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/host-continuation.md).
353
+
354
+ ---
355
+
356
+ ## Essential CLI Reference
357
+
358
+ | Goal | Command |
359
+ |---|---|
360
+ | **Check installation & 13-dimension health** | `python .agents/harness.py doctor [--install-check] [--json]` |
361
+ | **Explore module/feature symbol graph** | `python .agents/harness.py graph --feature <name> --json` |
362
+ | **Resolve target file/symbol AST context** | `python .agents/harness.py task-context --file <path> [--since <digest>] --json` |
363
+ | **Inspect Hybrid Flash delegation strategy** | `python .agents/harness.py delegate-plan [--task-id <id>] --json` |
364
+ | **Check active task next action** | `python .agents/harness.py task status --task-id <id> --next` |
365
+ | **Surgical AST symbol outline / edit** | `python .agents/harness.py symbol <outline\|read\|replace-body\|insert-after> ...` |
366
+ | **Record persistent project note / rule** | `python .agents/harness.py context <note\|instruct\|drift-report> ...` |
367
+ | **Bridge context to a fresh chat** | `python .agents/harness.py task bridge [--save \| --bridge-id <id> \| --snooze]` |
368
+ | **Complete parallel V2 reviews in 1 turn** | `python .agents/harness.py review complete-all --task <id>` |
369
+ | **Run physical-device / Maestro / UI gates** | `python .agents/harness.py <device install-start \| maestro run \| roborazzi verify>` |
370
+ | **Self-heal task state / session forensics** | `python .agents/harness.py <task heal --task-id <id> \| diagnose-session --json>` |
371
+
372
+ ---
373
+
374
+ ## Clear Boundaries: What It Is & What It Isn't
375
+
376
+ | Best Fit For | Not Intended To Be |
377
+ |---|---|
378
+ | Production Android & KMM/KMP teams shipping daily with AI agents | A replacement for Android Studio, Gradle, or physical-device testing |
379
+ | Multi-module, mixed legacy/modern repos (`Java` + `Kotlin`, `XML` + `Compose`, `Swift`) | An autonomous "no-human-in-the-loop" auto-committer (`GIT_POLICY = never`) |
380
+ | Apps with `Room`, `Auth`, `Billing`, permissions, `ForegroundService`, or RTL localization | An OS kernel sandbox (hooks enforce boundaries at the AI host tool layer) |
381
+ | Engineers who want frontier AI speed **without** surrendering architectural control | Needed for 50-line throwaway prototypes or one-off snippets |
382
+
383
+ ---
384
+
385
+ ## Repository Architecture & CI Validation
386
+
387
+ ```text
388
+ agents/
389
+ ├── harness.py # Zero-overhead in-process CLI router & command dispatcher
390
+ ├── rules/ # Canonical English-only behavioral kernel (<= 36k combined budget)
391
+ ├── scripts/ # Pure Python 3.10–3.14 stdlib governance, AST graph, gates & doctor
392
+ ├── skills/ # On-demand domain skills (efficient-context, phone-e2e, UI diagnostics)
393
+ ├── subagents/ # Epistemically isolated V2 reviewer & specialist templates
394
+ ├── tool-adapters/ # Managed templates for Antigravity, Claude Code, Cursor, Codex & more
395
+ └── workflows/ # Multi-phase, phone-e2e, Android inspection & PM tracker playbooks
396
+ ```
397
+
398
+ - **Cross-Platform CI Matrix**: Validated on **Ubuntu 24.04**, **Windows Server 2025**, and **macOS 15** across **Python 3.10–3.14**, including `37` deterministic selftest suites (`python harness_cli.py selftest`), wheel packaging lifecycle tests, release checksum verification (`263` tracked files), and adversarial hook/security tests.
399
+ - **Contributing**: Runtime code uses **Python standard library only** (`0` third-party runtime dependencies) and preserves all approval, nonce, and snapshot-binding invariants. See [CONTRIBUTING.md](https://github.com/rabee-elkholy/android-agent-harness/blob/main/CONTRIBUTING.md) and [SECURITY.md](https://github.com/rabee-elkholy/android-agent-harness/blob/main/SECURITY.md).
400
+
401
+ ---
402
+
403
+ ## FAQ
404
+
405
+ <details>
406
+ <summary><b>1. Does this make the AI agent fully autonomous or allow it to run git commit / push?</b></summary>
407
+ <br />
408
+ No. Android Agent Harness is strictly <b>approval-first</b> and enforces <b>Human Git Authority</b>. The agent must obtain explicit developer approval on the plan before mutating source files, and client installations leave <code>git commit</code>, <code>branch</code>, <code>merge</code>, <code>rebase</code>, and <code>push</code> exclusively to the developer.
409
+ </details>
410
+
411
+ <details>
412
+ <summary><b>2. Does every task spawn multiple AI reviewers or require a physical phone?</b></summary>
413
+ <br />
414
+ No. The 6-tier risk engine routes verification from the actual diff: <code>T0</code> docs and <code>T1</code> resource tweaks skip AI reviewer subagents completely, isolated <code>&lt;=25</code>-line private Kotlin edits consolidate to a single logic/test reviewer, and physical-device verification runs only when device/UI/Room surfaces warrant it (or when requested).
415
+ </details>
416
+
417
+ <details>
418
+ <summary><b>3. Does it work with KMM / KMP (Kotlin Multiplatform) and mixed legacy Java + XML projects?</b></summary>
419
+ <br />
420
+ Yes. It indexes <code>Kotlin</code>, <code>Java</code>, <code>XML</code>, and iOS <code>Swift</code> across <code>commonMain</code>, <code>androidMain</code>, <code>iosMain</code>, <code>:shared</code>, and <code>:composeApp</code>, enforces <code>commonMain</code> platform purity and <code>expect</code>/<code>actual</code> pairing, and preserves local module architectures (<code>MVP</code>, <code>MVVM</code>, <code>MVI</code>) without forcing migrations.
421
+ </details>
422
+
423
+ <details>
424
+ <summary><b>4. Can I keep multiple tasks open in the same worktree without committing first?</b></summary>
425
+ <br />
426
+ Yes — up to <b>three live tasks</b> may share a single worktree simultaneously as long as their planned file sets are disjoint. Switch active tasks with <code>python .agents/harness.py task switch --task-id &lt;id&gt;</code>, preview/combine pre-approval plans with <code>task merge</code>, and finalize individual developer commits via <code>task reconcile-delivery --task-id &lt;id&gt;</code>.
427
+ </details>
428
+
429
+ <details>
430
+ <summary><b>5. How do I maximize weekly frontier model quota (e.g., Google AI Ultra) on large repositories?</b></summary>
431
+ <br />
432
+ Keep your main Antigravity chat on <code>Gemini 4 Carbon (Medium)</code> for architecture, planning, and <code>T4/T5</code> sensitive edits, and rotate to <code>Claude Sonnet 5.5</code> / <code>Opus 5.5</code> when <code>/usage</code> approaches <code>~50%</code> (separate quota pool; <code>task bridge</code> and <code>task continue</code> preserve full task state). Meanwhile, pinned <code>gemini-3.8-flash-medium</code> <code>harness-scout-agent</code> and <code>harness-worker-agent</code> absorb discovery sweeps, <code>T0–T3</code> implementation slices, and Gradle compile/test loops.
433
+ </details>
434
+
435
+ <details>
436
+ <summary><b>6. What happens if managed files under .agents/ are accidentally edited or corrupted?</b></summary>
437
+ <br />
438
+ Run <code>python .agents/harness.py doctor</code> to diagnose any drift across 13 health dimensions, and <code>python ~/.android-harness/kit/harness_cli.py repair --repo /path/to/app</code> to restore checksum-verified engine files without touching your application source code or project notes.
439
+ </details>
440
+
441
+ ---
442
+
443
+ ## Documentation Hub
444
+
445
+ - [Quickstart Guide](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/quickstart.md) · [Install or Update Prompt](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/install-or-update-prompt.md) · [Setup Wizard Reference](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/setup-wizard.md)
446
+ - [Architecture Reference](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/architecture.md) · [Workflow Guide](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/workflows.md) · [Efficient Task Context](https://github.com/rabee-elkholy/android-agent-harness/blob/main/agents/workflows/efficient-context.md)
447
+ - [Reviewer Model Configuration](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/reviewer-models.md) · [Verification Value & Metrics](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/verification-value.md) · [PM MCP Integrations](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/workflows/pm-integrations.md)
448
+ - [Compatibility Matrix](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/compatibility-matrix.md) · [AI Tool Support](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/tool-support.md) · [Cross-Host Continuation](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/host-continuation.md)
449
+ - [Threat Model](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/threat-model.md) · [Restore & Recovery](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/restore.md) · [v1.3.2 Engine & Context Benchmarks](https://github.com/rabee-elkholy/android-agent-harness/blob/main/docs/benchmark/v1.3.2-engine-and-context-benchmarks.md)
450
+
451
+ ---
452
+
453
+ ## License
454
+
455
+ MIT License. See [LICENSE](https://github.com/rabee-elkholy/android-agent-harness/blob/main/LICENSE).