argus-testing 0.13.0__tar.gz → 0.13.2__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 (219) hide show
  1. {argus_testing-0.13.0 → argus_testing-0.13.2}/AGENTS.md +44 -8
  2. {argus_testing-0.13.0 → argus_testing-0.13.2}/CHANGELOG.md +119 -0
  3. {argus_testing-0.13.0 → argus_testing-0.13.2}/PKG-INFO +19 -18
  4. {argus_testing-0.13.0 → argus_testing-0.13.2}/PRIVACY.md +3 -0
  5. {argus_testing-0.13.0 → argus_testing-0.13.2}/README.md +18 -17
  6. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/reporting/html.py +52 -3
  7. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/reporting/json.py +4 -0
  8. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/reporting/junit.py +36 -1
  9. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/reporting/sarif.py +46 -0
  10. argus_testing-0.13.2/argus/adapters/scripts/scan_step_js.js +39 -0
  11. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/capsules.py +13 -6
  12. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/coverage.py +2 -1
  13. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/evidence.py +14 -0
  14. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/findings.py +73 -17
  15. argus_testing-0.13.2/argus/application/finish.py +268 -0
  16. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/lifecycle.py +45 -6
  17. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/manager.py +7 -2
  18. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/navigation.py +4 -1
  19. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/probes.py +25 -3
  20. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/progress.py +6 -0
  21. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/regression.py +62 -20
  22. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/session.py +5 -0
  23. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/testgen.py +53 -20
  24. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/text.py +29 -22
  25. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/verification.py +2 -2
  26. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/browser.py +38 -0
  27. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/doctor.py +89 -3
  28. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/domain/models.py +1 -0
  29. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/domain/verdict.py +44 -13
  30. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/entrypoints/mcp/results.py +9 -1
  31. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/entrypoints/mcp/server.py +11 -4
  32. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/mcp_server.py +1 -1
  33. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/settings.py +3 -0
  34. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/gateway.py +1 -0
  35. {argus_testing-0.13.0 → argus_testing-0.13.2}/docs/agent-protocol.md +84 -13
  36. {argus_testing-0.13.0 → argus_testing-0.13.2}/docs/evaluation.md +145 -2
  37. {argus_testing-0.13.0 → argus_testing-0.13.2}/docs/usage.md +31 -14
  38. {argus_testing-0.13.0 → argus_testing-0.13.2}/examples/response-overhead.json +7 -7
  39. {argus_testing-0.13.0 → argus_testing-0.13.2}/llms-install.md +8 -3
  40. {argus_testing-0.13.0 → argus_testing-0.13.2}/pyproject.toml +1 -1
  41. {argus_testing-0.13.0 → argus_testing-0.13.2}/server.json +2 -2
  42. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_agent_results.py +30 -3
  43. argus_testing-0.13.2/tests/test_autonomous_acceptance.py +664 -0
  44. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_codex_evaluation.py +5 -1
  45. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_doctor.py +48 -0
  46. argus_testing-0.13.2/tests/test_first_run_browser.py +119 -0
  47. argus_testing-0.13.2/tests/test_regression_reports.py +271 -0
  48. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_stdio_protocol.py +43 -3
  49. argus_testing-0.13.2/tests/test_structured_results.py +179 -0
  50. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_verdict.py +8 -5
  51. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_verdict_integrity.py +2 -1
  52. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_verdict_parity.py +115 -2
  53. argus_testing-0.13.2/tests/test_verdict_properties.py +181 -0
  54. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_verification_efficiency.py +23 -5
  55. argus_testing-0.13.0/argus/application/finish.py +0 -152
  56. {argus_testing-0.13.0 → argus_testing-0.13.2}/.gitignore +0 -0
  57. {argus_testing-0.13.0 → argus_testing-0.13.2}/CONTRIBUTING.md +0 -0
  58. {argus_testing-0.13.0 → argus_testing-0.13.2}/LICENSE +0 -0
  59. {argus_testing-0.13.0 → argus_testing-0.13.2}/SECURITY.md +0 -0
  60. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/__init__.py +0 -0
  61. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/_compat.py +0 -0
  62. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/access.py +0 -0
  63. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/__init__.py +0 -0
  64. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/assets/__init__.py +0 -0
  65. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/assets/dashboard.css +0 -0
  66. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/assets/dashboard.html +0 -0
  67. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/assets/dashboard.js +0 -0
  68. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/journal.py +0 -0
  69. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/logging.py +0 -0
  70. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/monitoring.py +0 -0
  71. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/reporting/__init__.py +0 -0
  72. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/reporting/common.py +0 -0
  73. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/reporting/privacy.py +0 -0
  74. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/reporting/reporter.py +0 -0
  75. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/scripts/__init__.py +0 -0
  76. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/scripts/check_layout_js.js +0 -0
  77. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/scripts/collect_text_js.js +0 -0
  78. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/scripts/extract_elements_js.js +0 -0
  79. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/scripts/extract_page_content_js.js +0 -0
  80. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/scripts/extract_performance_js.js +0 -0
  81. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/scripts/find_visible_element_js.js +0 -0
  82. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/scripts/inspect_element_js.js +0 -0
  83. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/scripts/ref_engine_js.js +0 -0
  84. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/adapters/scripts/row_text_js.js +0 -0
  85. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/__init__.py +0 -0
  86. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/actions.py +0 -0
  87. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/capture.py +0 -0
  88. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/context.py +0 -0
  89. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/events.py +0 -0
  90. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/expectations.py +0 -0
  91. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/exports.py +0 -0
  92. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/guidance.py +0 -0
  93. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/inspection.py +0 -0
  94. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/journeys.py +0 -0
  95. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/network_storage.py +0 -0
  96. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/observation.py +0 -0
  97. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/operations.py +0 -0
  98. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/payloads.py +0 -0
  99. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/screen.py +0 -0
  100. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/screenshots.py +0 -0
  101. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/telemetry.py +0 -0
  102. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/application/verifiers.py +0 -0
  103. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/bootstrap.py +0 -0
  104. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/cleanup.py +0 -0
  105. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/cli.py +0 -0
  106. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/detector.py +0 -0
  107. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/differ.py +0 -0
  108. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/domain/__init__.py +0 -0
  109. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/domain/contracts.py +0 -0
  110. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/domain/coverage.py +0 -0
  111. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/domain/detector.py +0 -0
  112. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/domain/findings.py +0 -0
  113. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/domain/replay.py +0 -0
  114. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/domain/verification.py +0 -0
  115. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/domain/wordchars.py +0 -0
  116. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/entrypoints/__init__.py +0 -0
  117. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/entrypoints/mcp/__init__.py +0 -0
  118. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/entrypoints/mcp/catalog.py +0 -0
  119. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/entrypoints/mcp/compat.py +0 -0
  120. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/entrypoints/mcp/overhead.py +0 -0
  121. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/entrypoints/mcp/snapshots.py +0 -0
  122. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/install_check.py +0 -0
  123. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/measurement.py +0 -0
  124. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/models.py +0 -0
  125. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/ports.py +0 -0
  126. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/reporter.py +0 -0
  127. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/resolver.py +0 -0
  128. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/runtime.py +0 -0
  129. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/screen/__init__.py +0 -0
  130. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/screen/backend.py +0 -0
  131. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/screen/permissions.py +0 -0
  132. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/screen/safety.py +0 -0
  133. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/screen/validate.py +0 -0
  134. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/security.py +0 -0
  135. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus/storage.py +0 -0
  136. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/__init__.py +0 -0
  137. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/bench/__init__.py +0 -0
  138. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/bench/__main__.py +0 -0
  139. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/bench/runner.py +0 -0
  140. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/bench/scenarios_buggytasks.py +0 -0
  141. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/bench/scenarios_darkshop.py +0 -0
  142. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/__init__.py +0 -0
  143. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/__main__.py +0 -0
  144. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/adjudication.py +0 -0
  145. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/agent_fixture.py +0 -0
  146. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/browser_baseline.py +0 -0
  147. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/codex_runner.py +0 -0
  148. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/fixture.py +0 -0
  149. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/models.py +0 -0
  150. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/overhead.py +0 -0
  151. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/scoring.py +0 -0
  152. {argus_testing-0.13.0 → argus_testing-0.13.2}/argus_lab/evaluation/smoke.py +0 -0
  153. {argus_testing-0.13.0 → argus_testing-0.13.2}/docs/architecture.md +0 -0
  154. {argus_testing-0.13.0 → argus_testing-0.13.2}/docs/decisions.md +0 -0
  155. {argus_testing-0.13.0 → argus_testing-0.13.2}/docs/evaluation-0.8.md +0 -0
  156. {argus_testing-0.13.0 → argus_testing-0.13.2}/examples/agent-evaluation-optimized-evidence.json +0 -0
  157. {argus_testing-0.13.0 → argus_testing-0.13.2}/examples/agent-evaluation-optimized.json +0 -0
  158. {argus_testing-0.13.0 → argus_testing-0.13.2}/examples/agent-evaluation-pilot-evidence.json +0 -0
  159. {argus_testing-0.13.0 → argus_testing-0.13.2}/examples/agent-evaluation-pilot.json +0 -0
  160. {argus_testing-0.13.0 → argus_testing-0.13.2}/examples/mcp-config.json +0 -0
  161. {argus_testing-0.13.0 → argus_testing-0.13.2}/glama.json +0 -0
  162. {argus_testing-0.13.0 → argus_testing-0.13.2}/human-eye-fixture/app.py +0 -0
  163. {argus_testing-0.13.0 → argus_testing-0.13.2}/scripts/check_distributions.py +0 -0
  164. {argus_testing-0.13.0 → argus_testing-0.13.2}/test-site/app.py +0 -0
  165. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/__init__.py +0 -0
  166. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/conftest.py +0 -0
  167. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_action_targeting.py +0 -0
  168. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_action_time_resolution.py +0 -0
  169. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_adjudication.py +0 -0
  170. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_agent_workflow.py +0 -0
  171. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_architecture_runtime.py +0 -0
  172. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_audit_hardening.py +0 -0
  173. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_bench_report.py +0 -0
  174. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_browser_shadow.py +0 -0
  175. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_crawl_site.py +0 -0
  176. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_cross_stack.py +0 -0
  177. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_deprecated_modules.py +0 -0
  178. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_detector_console.py +0 -0
  179. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_detector_network.py +0 -0
  180. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_distribution_contents.py +0 -0
  181. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_element_identity.py +0 -0
  182. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_evaluation.py +0 -0
  183. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_evidence_provenance.py +0 -0
  184. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_expectation_oracle.py +0 -0
  185. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_export_guards.py +0 -0
  186. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_export_privacy.py +0 -0
  187. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_export_test.py +0 -0
  188. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_final_coverage.py +0 -0
  189. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_final_hardening.py +0 -0
  190. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_form_checkboxes.py +0 -0
  191. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_frame_extraction.py +0 -0
  192. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_journal_regression.py +0 -0
  193. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_mcp_compat.py +0 -0
  194. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_mcp_contract.py +0 -0
  195. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_monitoring.py +0 -0
  196. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_real_app_interactions.py +0 -0
  197. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_regression_cli.py +0 -0
  198. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_release_metadata.py +0 -0
  199. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_replay_minimize.py +0 -0
  200. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_replay_receipt.py +0 -0
  201. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_replay_trace.py +0 -0
  202. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_reporter_badge.py +0 -0
  203. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_reproduction_receipt.py +0 -0
  204. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_resolver.py +0 -0
  205. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_response_efficiency.py +0 -0
  206. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_row_textarea.py +0 -0
  207. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_runtime_limits.py +0 -0
  208. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_screen_errors.py +0 -0
  209. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_screen_input_guard.py +0 -0
  210. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_screen_permissions.py +0 -0
  211. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_screen_safety.py +0 -0
  212. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_security_boundaries.py +0 -0
  213. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_session_lifecycle.py +0 -0
  214. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_state_capsules.py +0 -0
  215. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_storage_cleanup.py +0 -0
  216. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_substrate.py +0 -0
  217. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_textarea_label.py +0 -0
  218. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_verification_contract.py +0 -0
  219. {argus_testing-0.13.0 → argus_testing-0.13.2}/tests/test_visual_inspection.py +0 -0
@@ -21,7 +21,7 @@ brew install cliclick
21
21
  claude mcp add argus -- argus-mcp
22
22
  argus-mcp --version # confirm host loaded the version you expect
23
23
  argus-mcp --list-tools # show the selected profile and public tools
24
- argus-mcp --doctor # Chromium, report dir; on macOS also screen grants
24
+ argus-mcp --doctor # Chromium and report dir decide the result; screen grants are listed
25
25
  ```
26
26
 
27
27
  The default `web` profile has 29 tools covering the complete testing loop,
@@ -60,15 +60,45 @@ the user's broader task. The short version:
60
60
  Map → **Use it** (walk each goal end-to-end, carrying real state) →
61
61
  Hypothesize → Act → Observe → Verify → Record → Cover.
62
62
 
63
+ ## Running a pass without the user
64
+
65
+ The user wants the result, not the run. Handle the routine parts yourself and
66
+ ask only for what you cannot get: a login you were not given, or a decision
67
+ outside the task.
68
+
69
+ - **Address**: from the request, the conversation, the project's start script
70
+ and port, or a dev server that is already running. Start the app with the
71
+ project's own command when that is part of your task.
72
+ - **Setup**: the first `start_session` downloads the browser; if it reports the
73
+ download is still running, wait and call it again.
74
+ - **A target is gone or ambiguous**: `observe` and choose from what is there now.
75
+ - **A timeout or error after a write**: check whether the write happened
76
+ (`observe`, `verify_persistence`) before repeating it.
77
+ - **A failed session**: call `start_session` again with the unfinished goals.
78
+ It saves the earlier session's report first and returns it as
79
+ `data.unended_run`; confirmed findings are already in the journal.
80
+ - **No account for a journey**: mark the goal `blocked` with the reason, finish
81
+ the rest, and say what you need in the report.
82
+ - **After a fix**: start a session and call `regression_check`. When the bug
83
+ was a change that did not stay, repeat that action once on the fixed build
84
+ first, because the old data is still there.
85
+ - **A test file the user asked for**: call `export_test` before `end_session`.
86
+ It reads the open session's findings; a closed session has none to export.
87
+ - **Report**: the journeys tested, each bug with impact, steps and evidence
88
+ paths, what was not finished and why. Inconclusive stays inconclusive, and a
89
+ run with no defects is a normal result.
90
+
63
91
  ## Structured results (0.8)
64
92
 
65
93
  Read `structuredContent.data`
66
94
  for operation-specific evidence where available; text and images remain for
67
95
  compatibility. `status: completed` means the tool ran, not that the application
68
96
  passed. Inspect expectation matches, finding verdicts and export warnings.
69
- Progress notifications describe actual phases, not percentages or proof of
70
- coverage. Missing live notifications do not mean execution stopped; the final
71
- result includes captured stages. See [the protocol](docs/agent-protocol.md).
97
+ Since 0.13.2, MCP progress notifications are off unless
98
+ `ARGUS_PROGRESS_NOTIFICATIONS=1`; every result lists its `stages`. When on, they
99
+ describe actual phases, not percentages or proof of coverage, and missing
100
+ notifications do not mean execution stopped. See
101
+ [the protocol](docs/agent-protocol.md#progress-transport).
72
102
 
73
103
  ## Tools you'll use most
74
104
 
@@ -84,7 +114,7 @@ result includes captured stages. See [the protocol](docs/agent-protocol.md).
84
114
  | `test_action(target, expect=...)` | Click + before/after diff in one call. Pass `expect` to PREDICT the outcome ({"count":{"label":"tasks","delta":1}}, {"gains":"Buy milk"}, {"removes":...}, {"text_present":...}, {"toast":...}, {"url_changed":true}) and Argus reports MATCH / SURPRISE — a surprise is a bug lead. Also shows CROSS-STACK: which requests the click fired (methods/statuses) and a CHECK nudge when a message appeared without a matching write. |
85
115
  | `verify_persistence(expect, target_text, after_url)` | Forces a fresh GET; returns target presence and the page observation. Check persistence once, repeat only to resolve uncertainty. A matching symptom is not proof of a semantic defect. |
86
116
  | `capsule_save(name, liveness_marker)` / `capsule_restore(name)` | Snapshot cookies and page storage after establishing state through the UI. Restore replaces the captured client state and checks a saved liveness marker when provided; without one, validity remains unverified. It does not reset backend data or guarantee identical application state across journeys. |
87
- | `regression_check()` | Re-test journaled findings against the current build: `STILL-PRESENT`, `FIXED` (the recorded symptom no longer reproduces), or `INCONCLUSIVE`. Mixed receipts remain inconclusive; a `FIXED` result is not proof of application correctness. |
117
+ | `regression_check()` | Re-test journaled findings against the current build: `STILL-PRESENT`, `NO-LONGER-REPRODUCES` or `INCONCLUSIVE` per finding, in `data.results` and the text. Mixed receipts remain inconclusive; no longer reproducing is not proof of application correctness. The `argus-regression` CLI reports the same outcome as `FIXED`. |
88
118
  | `export_test(finding_id, run=False, save_private_values=False)` | Export a standalone Playwright test from a finding verified by a fresh-load text check. Confirm it fails on the symptom before using it in CI. Unsupported steps are marked TODO; 0.12.1 blocks incomplete exports before execution. See the [export protocol](docs/agent-protocol.md#exporting-a-regression-test) for environment variables and package-version limits. `run=true` repeats the journey's writes, so use it only with the user's authority. Only set `save_private_values=true` when the user explicitly requests retaining local replay values; its returned JSON file can contain credentials and must never be published. |
89
119
  | `inspect_element(description)` | Computed styles + ARIA + outerHTML + truncation flag for interactive or visible non-interactive content. |
90
120
  | `check_layout()` | Bounded viewport signals for horizontal overflow, clipped text, small targets, and fixed/sticky overlays. |
@@ -137,7 +167,10 @@ goal evidence, user constraints, page coverage, time-budget status, review mode,
137
167
  observations, tool-call counts, step counts, and screenshot metadata.
138
168
 
139
169
  Since 0.10.0, persistence mismatches may include `suggested_verify` for
140
- a fresh-load recheck without repeating writes. Do not use replay for an
170
+ a fresh-load recheck without repeating writes. Since 0.13.2, a finding recorded
171
+ on that page without a confirmation also comes back with that
172
+ `suggested_verify`, offered and not applied; pass it with the `finding_id` only
173
+ if the finding is that failure. Do not use replay for an
141
174
  unrecorded refresh. Supplement an existing finding with
142
175
  `record_bug(title, severity, verify=..., finding_id=...)`, omitting evidence and
143
176
  keeping title/severity unchanged. At most two rechecks are allowed; stop on a
@@ -200,8 +233,11 @@ Don't record speculation, polish nits, or static a11y / SEO that
200
233
 
201
234
  - `argus-mcp --version` — what version your MCP host actually loaded.
202
235
  - `argus-mcp --doctor` — checks the matching Chromium (with a real headless
203
- launch) and the report directory, printing exact fixes; on macOS it also
204
- probes Screen Recording and Accessibility for screen mode.
236
+ launch) and the report directory, printing exact fixes. Those web checks
237
+ decide its exit status; a browser that is not downloaded yet is reported as
238
+ arriving on first use, not as a failure. On macOS it also lists the Screen
239
+ Recording and Accessibility grants, which count toward the status only with
240
+ `--tool-profile screen` or `full`.
205
241
  - `argus-mcp --list-tools` — selected tool profile and exported names.
206
242
  - `argus-mcp --tool-profile web|advanced|screen|full` — control discovery size.
207
243
  - `argus-mcp --unsafe` — turn on `eval_js` (off by default).
@@ -1,5 +1,124 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.13.2 - 2026-11-18
4
+
5
+ Fixes found by running Argus under a real coding agent and by an independent
6
+ review of that work. No new tools. Upgrade notes: absence that Argus could not
7
+ fully check is now inconclusive where it used to be confirmed, so a regression
8
+ gate can report more inconclusive results; exported tests carry verdict rules
9
+ version 2 only after they are exported again; MCP progress notifications are
10
+ off unless `ARGUS_PROGRESS_NOTIFICATIONS=1`.
11
+
12
+ - A fresh load that kept the path but dropped or changed a requested query
13
+ parameter (`/tasks?list=work` redirected to `/tasks?list=home`) is a
14
+ different page: text absent there is inconclusive (`moved`) instead of a
15
+ confirmed absence. Added parameters, parameter order and tracking parameters
16
+ (`utm_*`, click ids) do not count.
17
+ - The search for text a lazy or virtualized list has not rendered now scrolls
18
+ in steps shorter than what the scroller shows (a 1000px step skipped rows of
19
+ a list without overscan), also scrolls boxes with their own scrollbar, of
20
+ any height, and the lists inside iframes (nested and cross-origin ones
21
+ included), and reports whether it reached the end. A search that spends its 30 steps with content
22
+ left is inconclusive (`unsearched`) instead of a confirmed absence. Live
23
+ verification, `verify_persistence` and exported tests (verdict rules
24
+ version 2) share the change; export a test again to pick it up.
25
+ - `argus-mcp --doctor` exits 0 when web testing can run. Missing macOS screen
26
+ grants are listed but only fail the check with `--tool-profile screen` or
27
+ `full`, and a browser that is not downloaded yet is reported as arriving on
28
+ the first session instead of as missing.
29
+ - A session that was never ended no longer loses what it found. A confirmed
30
+ finding is journaled when `record_bug` confirms it, not at `end_session`.
31
+ `start_session` after a timeout or failure first saves the earlier session's
32
+ report and names it, with the goals that were left, in its result
33
+ (`unended_run`). The timeout and failed-session errors say how to continue
34
+ and to check whether a write happened before repeating it.
35
+ - `navigate` accepts a path such as `/settings`, resolved against the current
36
+ page, as `verify_persistence` and `record_bug` already did.
37
+ - MCP progress notifications are off unless `ARGUS_PROGRESS_NOTIFICATIONS=1`,
38
+ and the closing stage is never sent. In a real Claude Code session the
39
+ server was restarted twice in one run, losing the session and its findings:
40
+ a progress notification read together with its response reaches the host
41
+ for a token it has already dropped, and the host closes the connection.
42
+ Results still list their `stages`.
43
+ - After a failed `verify_persistence`, a finding recorded on that page without
44
+ a confirmation carries `suggested_verify`: the symptom that check saw,
45
+ offered and not applied. The caller confirms the finding with a second
46
+ `record_bug` call using its `finding_id` and that clause. `record_bug` does
47
+ not derive a condition from an earlier check or from a finding's wording,
48
+ and does not reverse one the caller stated; a condition given in `verify`
49
+ or in the evidence fields is checked as given. A `goal` named in a session that declared none is
50
+ dropped with a warning instead of rejecting the finding.
51
+ - `docs/evaluation.md` records two small host-level runs (Claude Code with a
52
+ small model, one synthetic to-do app with a seeded defect): what finished
53
+ without a person, what did not use Argus at all, the defects they found in
54
+ Argus, and what they cannot show. They are not a success rate.
55
+ - The server instructions and the session protocol tell a host to find the
56
+ app's address in the project before asking, and to report before changing
57
+ code.
58
+ - The protocol `start_session` returns has a seventh step on recovering
59
+ without the user, so a host connected without a plugin gets it too.
60
+ `AGENTS.md` and `docs/evaluation.md` describe running a pass unattended and
61
+ how to accept it: deterministic scenarios in CI, and a host-level protocol
62
+ with defective, fixed and never-defective builds.
63
+ - A host that reads only the structured result now gets every answer. Claude
64
+ Code shows the model the structured envelope and not the text, and tools
65
+ that answered in text, `regression_check` and `get_errors` among them,
66
+ reached it as `data: {}`: a real host called `regression_check`, saw
67
+ nothing, and re-tested by hand. `regression_check` now returns its results
68
+ in `data` (status per finding, counts, the reason when nothing is
69
+ journaled), and any other text-only result or error carries its text,
70
+ redacted, in `data.message`.
71
+ - `capsule_save`, `capsule_restore` and `regression_check` report their
72
+ failures (no such capsule, a write that failed, not a web session) as
73
+ errors instead of completed results.
74
+ - `regression_check` results now reach the report. A real host re-tested a
75
+ fix and handed the user a report with no recheck in it. `end_session` and
76
+ the HTML, JSON, JUnit and SARIF reports now show each journaled finding's
77
+ latest status (still present, no longer reproduces, inconclusive) with the
78
+ earlier ones, and a call with nothing journaled as such. They are kept
79
+ apart from new findings and are not journaled again. JUnit and SARIF never
80
+ count inconclusive as a pass. An HTML report with rechecks says "No new
81
+ findings recorded in this session" instead of "No bugs found".
82
+
83
+ - Both plugin test skills, in the next plugin update: to find the app's
84
+ address, read the README and start scripts first and never print or read a
85
+ whole `.env`; if nothing else gives the address, look up only the host or
86
+ port variable. A real host had run `cat .env` for this. This is guidance to
87
+ the agent; it cannot stop a host from reading a file.
88
+
89
+ ## Plugin updates - 2026-11-17 (0.13.1)
90
+
91
+ - Claude plugin 0.13.1 and Codex plugin 0.13.1 pin the published Argus 0.13.1
92
+ runtime and refresh their dependency locks and setup commands.
93
+ - Both test skills are rewritten so the agent can run a pass without handing
94
+ work back to the user. They say to find the address from the conversation,
95
+ the project or a running dev server instead of asking; let the first session download the browser and retry
96
+ while it does; say what to do when a target is gone, a page is slow, a call
97
+ times out after a write, the session fails, a login is missing or a check is
98
+ inconclusive; and report the journeys tested, each bug with its evidence,
99
+ and what was not finished. They also apply when a feature was just finished
100
+ and when a fix needs re-testing, which goes through the journal with
101
+ `regression_check`. After a run with a small model as the host, they also
102
+ say: look before asking, `verify` states the symptom and not the expected
103
+ state, and a bug found is recorded and reported, not fixed, unless a fix was
104
+ asked for. Exporting a test file is on request, not part of every
105
+ run, and happens before `end_session`, because `export_test` reads the open
106
+ session's findings. The skills no longer ask the user to set replay
107
+ variables.
108
+
109
+ ## 0.13.1 - 2026-11-12
110
+
111
+ - The first session downloads the Chromium build it needs instead of failing
112
+ with an install command, so connecting Argus is the only setup step. A
113
+ download that outlasts the call keeps running and `start_session` asks for a
114
+ retry. `ARGUS_AUTO_INSTALL_BROWSER=0` restores the old behaviour.
115
+
116
+ ## Plugin updates - 2026-11-12
117
+
118
+ - Claude plugin 0.13.0 and Codex plugin 0.13.0 pin the published Argus 0.13.0
119
+ runtime and refresh their dependency locks and setup commands. The two
120
+ plugins now share the runtime's version number.
121
+
3
122
  ## Plugin updates - 2026-11-10
4
123
 
5
124
  - Claude plugin 0.12.4 and Codex plugin 0.12.3 pin the published Argus 0.12.3
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.5
2
2
  Name: argus-testing
3
- Version: 0.13.0
3
+ Version: 0.13.2
4
4
  Summary: Testing MCP for coding agents: exercise real user journeys, find bugs, and verify them with evidence and persistence checks
5
5
  Project-URL: Homepage, https://yichenwu.dev/projects/argus
6
6
  Project-URL: Documentation, https://github.com/chriswu727/argus#readme
@@ -103,15 +103,8 @@ browser state and other unsupported steps require manual translation.
103
103
 
104
104
  ## Quick start
105
105
 
106
- Requires **Python 3.11+**, Chromium and an MCP-capable coding agent.
107
- With [`uv`](https://docs.astral.sh/uv/):
108
-
109
- ```bash
110
- uvx --from argus-testing python -m playwright install chromium
111
- uvx --from argus-testing argus-mcp --doctor
112
- ```
113
-
114
- Connect it to **Claude Code**:
106
+ Requires **Python 3.11+**, [`uv`](https://docs.astral.sh/uv/) and an MCP-capable
107
+ coding agent. Connect it to **Claude Code**:
115
108
 
116
109
  ```bash
117
110
  claude mcp add argus -- uvx --from argus-testing argus-mcp
@@ -123,20 +116,28 @@ Or **Codex**:
123
116
  codex mcp add argus -- uvx --from argus-testing argus-mcp
124
117
  ```
125
118
 
126
- Then ask your agent to test your app. Use test accounts and data you can change:
119
+ Then ask your agent to test your app. The first session downloads Chromium
120
+ once (about 150 MB); `uvx --from argus-testing argus-mcp --doctor` checks the
121
+ setup ahead of time. Use test accounts and data you can change:
127
122
  replaying a journey can repeat its writes.
128
123
 
129
124
  [Cursor, other MCP clients and troubleshooting](docs/usage.md#quick-start) ·
130
125
  [Claude plugin](plugins/claude) · [Codex plugin](plugins/codex)
131
126
 
132
- **0.13.0**: live verification, `verify_persistence` and exported tests now
133
- decide with the same rules, checked against each other on every change. They
134
- read what a user can see, in every frame and open shadow root, and an exported
135
- test reports an inconclusive check instead of a false failure. Regenerate
136
- exported tests to get the shared rules; see the
137
- [journal migration steps](docs/usage.md#regression-gates-in-ci), the
127
+ **0.13.2**: fixes found by running Argus under a real coding agent. Argus no
128
+ longer sends the progress notifications that made Claude Code drop the
129
+ connection partway through a run and lose the session's findings. A delete
130
+ that redirects to another list, or a row
131
+ further down a long or virtualized list than Argus scrolled, is reported as
132
+ inconclusive instead of confirmed absent. A finding is confirmed only by a
133
+ check the agent states. Structured tool results now carry the answer and error
134
+ reason, and regression conclusions appear in the session summary and every
135
+ report. `--doctor` passes on a Mac without the screen-mode grants. Since 0.13.1 the first session downloads Chromium by itself, and since
136
+ 0.13.0 live verification, `verify_persistence` and exported tests decide with
137
+ the same rules. Regenerate exported tests to pick up verdict rules version 2;
138
+ see the [journal migration steps](docs/usage.md#regression-gates-in-ci), the
138
139
  [releases](https://github.com/chriswu727/argus/releases) and the
139
- [changelog](CHANGELOG.md#0130---2026-11-12).
140
+ [changelog](CHANGELOG.md#0132---2026-11-18).
140
141
 
141
142
  ## Engineering
142
143
 
@@ -79,6 +79,9 @@ does not sanitize the separate application evidence described above.
79
79
  an Argus-operated model service.
80
80
  - Installing or upgrading through PyPI, `pip`, or `uvx` contacts those package
81
81
  distribution services under their own policies.
82
+ - When the Chromium build Argus needs is missing, the first session downloads it
83
+ from Playwright's distribution hosts. Nothing about the tested app is sent.
84
+ `ARGUS_AUTO_INSTALL_BROWSER=0` turns this off.
82
85
  - The optional `argus_lab.evaluation.codex_runner` sends synthetic testing prompts
83
86
  and tool evidence to the operator's ChatGPT-backed Codex session only after
84
87
  explicit subscription-usage opt-in. It stores local answers, usage, tool traces
@@ -59,15 +59,8 @@ browser state and other unsupported steps require manual translation.
59
59
 
60
60
  ## Quick start
61
61
 
62
- Requires **Python 3.11+**, Chromium and an MCP-capable coding agent.
63
- With [`uv`](https://docs.astral.sh/uv/):
64
-
65
- ```bash
66
- uvx --from argus-testing python -m playwright install chromium
67
- uvx --from argus-testing argus-mcp --doctor
68
- ```
69
-
70
- Connect it to **Claude Code**:
62
+ Requires **Python 3.11+**, [`uv`](https://docs.astral.sh/uv/) and an MCP-capable
63
+ coding agent. Connect it to **Claude Code**:
71
64
 
72
65
  ```bash
73
66
  claude mcp add argus -- uvx --from argus-testing argus-mcp
@@ -79,20 +72,28 @@ Or **Codex**:
79
72
  codex mcp add argus -- uvx --from argus-testing argus-mcp
80
73
  ```
81
74
 
82
- Then ask your agent to test your app. Use test accounts and data you can change:
75
+ Then ask your agent to test your app. The first session downloads Chromium
76
+ once (about 150 MB); `uvx --from argus-testing argus-mcp --doctor` checks the
77
+ setup ahead of time. Use test accounts and data you can change:
83
78
  replaying a journey can repeat its writes.
84
79
 
85
80
  [Cursor, other MCP clients and troubleshooting](docs/usage.md#quick-start) ·
86
81
  [Claude plugin](plugins/claude) · [Codex plugin](plugins/codex)
87
82
 
88
- **0.13.0**: live verification, `verify_persistence` and exported tests now
89
- decide with the same rules, checked against each other on every change. They
90
- read what a user can see, in every frame and open shadow root, and an exported
91
- test reports an inconclusive check instead of a false failure. Regenerate
92
- exported tests to get the shared rules; see the
93
- [journal migration steps](docs/usage.md#regression-gates-in-ci), the
83
+ **0.13.2**: fixes found by running Argus under a real coding agent. Argus no
84
+ longer sends the progress notifications that made Claude Code drop the
85
+ connection partway through a run and lose the session's findings. A delete
86
+ that redirects to another list, or a row
87
+ further down a long or virtualized list than Argus scrolled, is reported as
88
+ inconclusive instead of confirmed absent. A finding is confirmed only by a
89
+ check the agent states. Structured tool results now carry the answer and error
90
+ reason, and regression conclusions appear in the session summary and every
91
+ report. `--doctor` passes on a Mac without the screen-mode grants. Since 0.13.1 the first session downloads Chromium by itself, and since
92
+ 0.13.0 live verification, `verify_persistence` and exported tests decide with
93
+ the same rules. Regenerate exported tests to pick up verdict rules version 2;
94
+ see the [journal migration steps](docs/usage.md#regression-gates-in-ci), the
94
95
  [releases](https://github.com/chriswu727/argus/releases) and the
95
- [changelog](CHANGELOG.md#0130---2026-11-12).
96
+ [changelog](CHANGELOG.md#0132---2026-11-18).
96
97
 
97
98
  ## Engineering
98
99
 
@@ -243,7 +243,16 @@ def render_html(
243
243
  )
244
244
  else:
245
245
  coverage_section = f"<h2>Focus Areas</h2><ul>{focus}</ul>"
246
- no_bugs = '<div class="nb">No bugs found.</div>' if not r.bugs else ""
246
+ # After a recheck, "no bugs" would contradict a known bug that is still
247
+ # present: say what is true, that this session recorded nothing new.
248
+ findings_label = "New Findings" if r.regression else "Bugs Found"
249
+ if r.bugs:
250
+ no_bugs = ""
251
+ elif r.regression:
252
+ # Neutral, not the green all-clear: a known bug may still be present.
253
+ no_bugs = '<p class="nn">No new findings recorded in this session.</p>'
254
+ else:
255
+ no_bugs = '<div class="nb">No bugs found.</div>'
247
256
 
248
257
  observations_html = ""
249
258
  if r.observations:
@@ -260,6 +269,38 @@ def render_html(
260
269
  <h3>{_esc(observation.title)}</h3>
261
270
  <p>{_esc(observation.evidence)}</p>{image}</div>"""
262
271
 
272
+ regression_html = ""
273
+ if r.regression and not r.regression["findings"]:
274
+ regression_html = (
275
+ "<h2>Regression Re-check (0)</h2>"
276
+ f'<p class="cb">{r.regression["checks"]} run(s): no journaled findings for this '
277
+ "site; nothing was rechecked.</p>"
278
+ )
279
+ elif r.regression:
280
+ status_style = {
281
+ "STILL-PRESENT": "cv-bad",
282
+ "NO-LONGER-REPRODUCES": "cv-ok",
283
+ "INCONCLUSIVE": "cv-unknown",
284
+ }
285
+ rows = "".join(
286
+ f'<tr><td><span class="cv {status_style.get(f["status"], "cv-unknown")}">'
287
+ f"{_esc(f['status'])}</span></td><td>{_esc(str(f['severity']).upper())}</td>"
288
+ f"<td>{_esc(f['title'])}</td><td>{_esc(f.get('runs') or '')}</td>"
289
+ f"<td>{_esc(f.get('reason') or '')}</td><td>{_esc(' → '.join(f['history']))}</td></tr>"
290
+ for f in r.regression["findings"]
291
+ )
292
+ counts = r.regression["counts"]
293
+ regression_html = (
294
+ f"<h2>Regression Re-check ({len(r.regression['findings'])})</h2>"
295
+ f'<p class="cb">Latest of {r.regression["checks"]} run(s): '
296
+ f"{counts['still_present']} still present, {counts['no_longer_reproduces']} no "
297
+ f"longer reproduce, {counts['inconclusive']} inconclusive. Rechecks of findings "
298
+ "recorded earlier; they are not counted as new bugs.</p>"
299
+ '<div class="rw"><table class="rt"><tr><th>Status</th><th>Severity</th>'
300
+ "<th>Finding</th><th>Runs</th><th>Reason</th><th>History</th></tr>"
301
+ f"{rows}</table></div>"
302
+ )
303
+
263
304
  # screenshots timeline
264
305
  screenshots_html = ""
265
306
  if r.screenshots:
@@ -328,6 +369,13 @@ h3{{font-size:1.1rem;margin:.5rem 0;color:#f1f5f9}}
328
369
  ul{{margin:.5rem 0 .5rem 1.5rem;color:#cbd5e1}}
329
370
  li{{margin:.2rem 0}}
330
371
  .nb{{text-align:center;padding:3rem;color:#22c55e;font-size:1.2rem}}
372
+ .nn{{color:#94a3b8;margin:.5rem 0 1rem}}
373
+ .rw{{max-width:100%;overflow-x:auto;margin:.75rem 0 1rem}}
374
+ .rt{{width:100%;border-collapse:collapse;font-size:.85rem}}
375
+ .rt th,.rt td{{text-align:left;vertical-align:top;padding:.5rem .75rem;border-bottom:1px solid #334155}}
376
+ .rt th{{color:#94a3b8;font-weight:600;white-space:nowrap}}
377
+ .rt td{{color:#cbd5e1;overflow-wrap:break-word}}
378
+ .rt td:first-child{{white-space:nowrap}}
331
379
  .tl{{display:grid;grid-template-columns:repeat(auto-fit,minmax(400px,1fr));gap:1rem}}
332
380
  .tc{{background:#1e293b;border:1px solid #334155;border-radius:8px;overflow:hidden}}
333
381
  .th{{padding:.75rem 1rem;border-bottom:1px solid #334155}}
@@ -339,7 +387,7 @@ li{{margin:.2rem 0}}
339
387
  <h1>Argus QA Report</h1>
340
388
  <div class="m">{_esc(r.url)} — {_esc(r.review_mode)} review — {r.timestamp.strftime("%Y-%m-%d %H:%M:%S")} — {r.duration_seconds:.1f}s</div>
341
389
  <div class="sg">
342
- <div class="s"><div class="sv2">{len(r.bugs)}</div><div class="sl">Bugs Found</div></div>
390
+ <div class="s"><div class="sv2">{len(r.bugs)}</div><div class="sl">{findings_label}</div></div>
343
391
  <div class="s"><div class="sv2">{len(r.observations)}</div><div class="sl">Observations</div></div>
344
392
  <div class="s"><div class="sv2">{r.tool_calls}</div><div class="sl">Tool Calls</div></div>
345
393
  <div class="s"><div class="sv2">{r.actions_taken}</div><div class="sl">Recorded Steps</div></div>
@@ -347,7 +395,8 @@ li{{margin:.2rem 0}}
347
395
  <div class="s"><div class="sv2">{len(r.screenshots)}</div><div class="sl">Screenshots</div></div>
348
396
  </div>
349
397
  {coverage_section}
350
- <h2>Bugs ({len(r.bugs)})</h2><div class="sm">{summary}</div>
398
+ {regression_html}
399
+ <h2>{"New Findings" if r.regression else "Bugs"} ({len(r.bugs)})</h2><div class="sm">{summary}</div>
351
400
  {cards}{no_bugs}
352
401
  {observations_html}
353
402
  {screenshots_html}
@@ -36,4 +36,8 @@ def render_json(r: ExplorationResult) -> str:
36
36
  "summary": {"total": len(findings), "verified": verified, "by_severity": by_sev},
37
37
  "findings": findings,
38
38
  }
39
+ if r.regression:
40
+ # Rechecks of journaled findings, not new findings: summary.total does
41
+ # not count them.
42
+ doc["regression"] = r.regression
39
43
  return json.dumps(doc, indent=2, ensure_ascii=False)
@@ -16,17 +16,32 @@ def render_junit(r: ExplorationResult) -> str:
16
16
  import xml.etree.ElementTree as ET
17
17
 
18
18
  findings = [b.to_dict() for b in r.bugs]
19
+ rechecks = (r.regression or {}).get("findings", [])
19
20
  failures = sum(1 for f in findings if receipt_outcome(f.get("reproduction")) is not False)
21
+ failures += sum(1 for f in rechecks if f["status"] == "STILL-PRESENT")
20
22
  suite = ET.Element(
21
23
  "testsuite",
22
24
  {
23
25
  "name": "argus",
24
- "tests": str(len(findings)),
26
+ "tests": str(len(findings) + len(rechecks)),
25
27
  "failures": str(failures),
26
28
  "hostname": r.url or "",
27
29
  "time": f"{r.duration_seconds or 0:.1f}",
28
30
  },
29
31
  )
32
+ if r.regression:
33
+ properties = ET.SubElement(suite, "properties")
34
+ ET.SubElement(
35
+ properties,
36
+ "property",
37
+ {"name": "regression_checks", "value": str(r.regression["checks"])},
38
+ )
39
+ if r.regression.get("reason"):
40
+ ET.SubElement(
41
+ properties,
42
+ "property",
43
+ {"name": "regression_reason", "value": r.regression["reason"]},
44
+ )
30
45
  for f in findings:
31
46
  verdict = receipt_outcome(f.get("reproduction"))
32
47
  tier = (
@@ -64,6 +79,26 @@ def render_junit(r: ExplorationResult) -> str:
64
79
  else:
65
80
  fail = ET.SubElement(tc, "failure", {"message": f["title"][:200], "type": f["type"]})
66
81
  fail.text = "\n".join(body_lines)
82
+ # The same mapping as argus-regression: still present fails, no longer
83
+ # reproducing passes, inconclusive is skipped (never a pass).
84
+ for f in rechecks:
85
+ tc = ET.SubElement(
86
+ suite,
87
+ "testcase",
88
+ {
89
+ "classname": f"argus.regression.{f['severity']}",
90
+ "name": f"[{f['status']}] {f['title']}"[:250],
91
+ },
92
+ )
93
+ if f["status"] == "STILL-PRESENT":
94
+ fail = ET.SubElement(
95
+ tc, "failure", {"message": "known bug still present", "type": "regression"}
96
+ )
97
+ fail.text = f"{f['title']}\nruns={f.get('runs')}\npath={f.get('path')}"
98
+ elif f["status"] == "INCONCLUSIVE":
99
+ ET.SubElement(
100
+ tc, "skipped", {"message": f"symptom check inconclusive: {f.get('reason') or ''}"}
101
+ )
67
102
  return '<?xml version="1.0" encoding="UTF-8"?>\n' + xml_safe(
68
103
  ET.tostring(suite, encoding="unicode")
69
104
  )
@@ -53,6 +53,51 @@ def render_sarif(r: ExplorationResult) -> str:
53
53
  },
54
54
  }
55
55
  )
56
+ # Rechecks of journaled findings, in SARIF's own terms: still present
57
+ # fails, no longer reproducing passes, and an inconclusive check is "open"
58
+ # (undetermined), never a pass. SARIF requires level "none" when kind is
59
+ # not "fail".
60
+ _KIND = {"STILL-PRESENT": "fail", "NO-LONGER-REPRODUCES": "pass", "INCONCLUSIVE": "open"}
61
+ _LEVEL_BY_NAME = {severity.value: level for severity, level in _LEVEL.items()}
62
+ regression = r.regression or {}
63
+ for f in regression.get("findings", []):
64
+ rules.setdefault(
65
+ "regression-recheck",
66
+ {
67
+ "id": "regression-recheck",
68
+ "name": "regression-recheck",
69
+ "shortDescription": {"text": "Fresh-load recheck of a journaled finding"},
70
+ },
71
+ )
72
+ kind = _KIND[f["status"]]
73
+ level = _LEVEL_BY_NAME.get(f["severity"], "error") if kind == "fail" else "none"
74
+ reason = f" — {f['reason']}" if f.get("reason") else ""
75
+ results.append(
76
+ {
77
+ "ruleId": "regression-recheck",
78
+ "kind": kind,
79
+ "level": level,
80
+ "message": {"text": f"[{f['status']}] {f['title']}{reason}"},
81
+ "locations": [
82
+ {"physicalLocation": {"artifactLocation": {"uri": f.get("path") or "/"}}}
83
+ ],
84
+ "properties": {
85
+ "status": f["status"],
86
+ "history": f["history"],
87
+ "runs": f.get("runs"),
88
+ "severity": f["severity"],
89
+ },
90
+ }
91
+ )
92
+ run_properties = (
93
+ {
94
+ "regression": {
95
+ key: regression[key] for key in ("checks", "counts", "reason") if key in regression
96
+ }
97
+ }
98
+ if regression
99
+ else {}
100
+ )
56
101
  doc = {
57
102
  "$schema": "https://json.schemastore.org/sarif-2.1.0.json",
58
103
  "version": "2.1.0",
@@ -66,6 +111,7 @@ def render_sarif(r: ExplorationResult) -> str:
66
111
  }
67
112
  },
68
113
  "results": results,
114
+ **({"properties": run_properties} if run_properties else {}),
69
115
  }
70
116
  ],
71
117
  }
@@ -0,0 +1,39 @@
1
+ () => {
2
+ // One step of the search for text a lazy or virtualized list has not
3
+ // rendered yet, in one frame. Scrolls the frame's window, then each
4
+ // scrollable box (open shadow roots included). Returns false once nothing
5
+ // in this frame is left to scroll. Evaluated in every rendered frame by
6
+ // Argus and by exported tests, so it must stay self-contained.
7
+ //
8
+ // A step is never longer than what is visible, or the rows in between
9
+ // would never be rendered: nine tenths of the height, at least one pixel.
10
+ const step = (size) => Math.min(size, Math.max(1, Math.floor(size * 0.9)));
11
+ const page = document.scrollingElement || document.documentElement;
12
+ const before = window.scrollY;
13
+ if (window.innerHeight >= 1 && window.innerHeight + before < page.scrollHeight - 2) {
14
+ // 'instant' because a page with smooth scrolling would not have moved yet.
15
+ window.scrollBy({ top: step(window.innerHeight), behavior: 'instant' });
16
+ if (window.scrollY !== before) return true;
17
+ }
18
+ const boxes = [];
19
+ const visit = (root) => {
20
+ for (const el of root.querySelectorAll('*')) {
21
+ if (el.shadowRoot) visit(el.shadowRoot);
22
+ // A textarea's or select's content is its value, which the collector
23
+ // reads whole. A box with no height shows nothing to scroll through.
24
+ if (el.tagName === 'TEXTAREA' || el.tagName === 'SELECT' || el.clientHeight < 1) continue;
25
+ if (el.scrollTop + el.clientHeight >= el.scrollHeight - 2) continue;
26
+ const style = getComputedStyle(el);
27
+ if (style.visibility === 'hidden') continue;
28
+ const flow = style.overflowY;
29
+ if (flow === 'auto' || flow === 'scroll' || flow === 'overlay') boxes.push(el);
30
+ }
31
+ };
32
+ visit(document);
33
+ for (const el of boxes) {
34
+ const top = el.scrollTop;
35
+ el.scrollTo({ top: top + step(el.clientHeight), behavior: 'instant' });
36
+ if (el.scrollTop !== top) return true;
37
+ }
38
+ return false;
39
+ }