argus-testing 0.12.0__tar.gz → 0.12.2__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.
- {argus_testing-0.12.0 → argus_testing-0.12.2}/.gitignore +1 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/AGENTS.md +2 -3
- {argus_testing-0.12.0 → argus_testing-0.12.2}/CHANGELOG.md +49 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/CONTRIBUTING.md +7 -7
- argus_testing-0.12.2/PKG-INFO +181 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/PRIVACY.md +29 -1
- argus_testing-0.12.2/README.md +137 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/evidence.py +21 -9
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/exports.py +73 -9
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/findings.py +10 -2
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/regression.py +18 -3
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/testgen.py +272 -128
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/verification.py +35 -6
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/verifiers.py +3 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/browser.py +5 -2
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/cli.py +2 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/domain/models.py +2 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/domain/verification.py +23 -2
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/entrypoints/mcp/catalog.py +1 -1
- {argus_testing-0.12.0 → argus_testing-0.12.2}/docs/agent-protocol.md +40 -7
- {argus_testing-0.12.0 → argus_testing-0.12.2}/docs/evaluation.md +1 -1
- argus_testing-0.12.0/README.md → argus_testing-0.12.2/docs/usage.md +63 -271
- {argus_testing-0.12.0 → argus_testing-0.12.2}/examples/response-overhead.json +10 -10
- {argus_testing-0.12.0 → argus_testing-0.12.2}/llms-install.md +3 -3
- {argus_testing-0.12.0 → argus_testing-0.12.2}/pyproject.toml +2 -2
- {argus_testing-0.12.0 → argus_testing-0.12.2}/scripts/check_distributions.py +1 -1
- {argus_testing-0.12.0 → argus_testing-0.12.2}/server.json +3 -3
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_distribution_contents.py +6 -4
- argus_testing-0.12.2/tests/test_export_guards.py +354 -0
- argus_testing-0.12.2/tests/test_export_privacy.py +320 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_export_test.py +81 -33
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_real_app_interactions.py +8 -1
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_regression_cli.py +8 -1
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_release_metadata.py +17 -35
- argus_testing-0.12.2/tests/test_verification_contract.py +398 -0
- argus_testing-0.12.0/PKG-INFO +0 -764
- argus_testing-0.12.0/README.zh-CN.md +0 -334
- argus_testing-0.12.0/tests/test_export_guards.py +0 -169
- {argus_testing-0.12.0 → argus_testing-0.12.2}/LICENSE +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/SECURITY.md +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/_compat.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/access.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/assets/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/assets/dashboard.css +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/assets/dashboard.html +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/assets/dashboard.js +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/journal.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/logging.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/monitoring.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/reporting/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/reporting/common.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/reporting/html.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/reporting/json.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/reporting/junit.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/reporting/privacy.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/reporting/reporter.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/reporting/sarif.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/scripts/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/scripts/check_layout_js.js +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/scripts/extract_elements_js.js +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/scripts/extract_page_content_js.js +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/scripts/extract_performance_js.js +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/scripts/find_visible_element_js.js +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/adapters/scripts/inspect_element_js.js +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/actions.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/capsules.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/capture.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/context.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/coverage.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/events.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/expectations.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/finish.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/guidance.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/inspection.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/journeys.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/lifecycle.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/manager.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/navigation.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/network_storage.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/observation.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/operations.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/payloads.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/probes.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/progress.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/screen.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/screenshots.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/session.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/telemetry.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/application/text.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/bootstrap.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/cleanup.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/detector.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/differ.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/doctor.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/domain/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/domain/contracts.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/domain/coverage.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/domain/detector.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/domain/findings.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/domain/replay.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/entrypoints/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/entrypoints/mcp/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/entrypoints/mcp/compat.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/entrypoints/mcp/overhead.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/entrypoints/mcp/results.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/entrypoints/mcp/server.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/entrypoints/mcp/snapshots.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/install_check.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/mcp_server.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/measurement.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/models.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/ports.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/reporter.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/resolver.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/runtime.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/screen/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/screen/backend.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/screen/permissions.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/screen/safety.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/screen/validate.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/security.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/settings.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus/storage.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/bench/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/bench/__main__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/bench/runner.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/bench/scenarios_buggytasks.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/bench/scenarios_darkshop.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/__main__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/adjudication.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/agent_fixture.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/browser_baseline.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/codex_runner.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/fixture.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/gateway.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/models.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/overhead.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/scoring.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/argus_lab/evaluation/smoke.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/docs/architecture.md +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/docs/decisions.md +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/docs/evaluation-0.8.md +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/examples/agent-evaluation-optimized-evidence.json +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/examples/agent-evaluation-optimized.json +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/examples/agent-evaluation-pilot-evidence.json +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/examples/agent-evaluation-pilot.json +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/examples/mcp-config.json +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/glama.json +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/human-eye-fixture/app.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/test-site/app.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/__init__.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/conftest.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_action_targeting.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_adjudication.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_agent_results.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_agent_workflow.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_architecture_runtime.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_bench_report.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_browser_shadow.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_codex_evaluation.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_cross_stack.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_deprecated_modules.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_detector_console.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_detector_network.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_doctor.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_evaluation.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_evidence_provenance.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_expectation_oracle.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_final_coverage.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_final_hardening.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_form_checkboxes.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_journal_regression.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_mcp_compat.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_mcp_contract.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_monitoring.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_replay_minimize.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_replay_receipt.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_replay_trace.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_reporter_badge.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_reproduction_receipt.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_resolver.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_response_efficiency.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_runtime_limits.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_screen_permissions.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_screen_safety.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_security_boundaries.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_session_lifecycle.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_state_capsules.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_stdio_protocol.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_storage_cleanup.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_substrate.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_verification_efficiency.py +0 -0
- {argus_testing-0.12.0 → argus_testing-0.12.2}/tests/test_visual_inspection.py +0 -0
|
@@ -1,7 +1,6 @@
|
|
|
1
1
|
# Argus — Agent Guide
|
|
2
2
|
|
|
3
|
-
This
|
|
4
|
-
Argus through MCP. Argus is a QA capability inside the host's current
|
|
3
|
+
This guide is for an agent using Argus through MCP. Argus is a QA capability inside the host's current
|
|
5
4
|
task; it does not replace the host agent's identity or authority.
|
|
6
5
|
|
|
7
6
|
## Setup
|
|
@@ -86,7 +85,7 @@ result includes captured stages. See [the protocol](docs/agent-protocol.md).
|
|
|
86
85
|
| `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. |
|
|
87
86
|
| `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. |
|
|
88
87
|
| `regression_check()` | Re-test the findings journaled in prior runs against the CURRENT build: STILL-PRESENT / NO-LONGER-REPRODUCES / INCONCLUSIVE. "Did my fix land, did anything come back?" |
|
|
89
|
-
| `export_test(finding_id, run=False)` |
|
|
88
|
+
| `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. |
|
|
90
89
|
| `inspect_element(description)` | Computed styles + ARIA + outerHTML + truncation flag for interactive or visible non-interactive content. |
|
|
91
90
|
| `check_layout()` | Bounded viewport signals for horizontal overflow, clipped text, small targets, and fixed/sticky overlays. |
|
|
92
91
|
| `screenshot(name, element="", full_page=False)` | Full viewport, full page, or a tight crop; waits for finite CSS transitions, then returns the image plus its absolute evidence path. |
|
|
@@ -1,5 +1,54 @@
|
|
|
1
1
|
# Changelog
|
|
2
2
|
|
|
3
|
+
## 0.12.2 - 2026-11-08
|
|
4
|
+
|
|
5
|
+
- Treat intermittent regression receipts as inconclusive instead of fixed.
|
|
6
|
+
- Retain full verification predicates and preconditions privately for journals
|
|
7
|
+
and exports; public receipt text remains a preview. Hash complete predicates
|
|
8
|
+
for deduplication, including URL queries and fragments.
|
|
9
|
+
- Require re-recording legacy journal entries whose verification conditions are
|
|
10
|
+
incomplete. Storage-reset checks remain excluded from automatic regression
|
|
11
|
+
and test export rather than silently dropping the reset.
|
|
12
|
+
- Detect unexpected password login forms while allowing recorded checks of a
|
|
13
|
+
login page itself. Stop ambiguous row selection before replaying its action;
|
|
14
|
+
positional fallbacks require manual translation.
|
|
15
|
+
- Bound exported network waits like live verification. Busy pages can still
|
|
16
|
+
provide positive symptom evidence; unsettled absence stays inconclusive.
|
|
17
|
+
|
|
18
|
+
## Plugin updates - 2026-11-05
|
|
19
|
+
|
|
20
|
+
- Claude plugin 0.12.2 and Codex plugin 0.12.1 pin the published Argus 0.12.1
|
|
21
|
+
runtime, refresh their dependency locks and document private replay values.
|
|
22
|
+
- Keep raw replay values out of shared artifacts; retention remains opt-in.
|
|
23
|
+
|
|
24
|
+
## 0.12.1 - 2026-11-05
|
|
25
|
+
|
|
26
|
+
- Keep all application URLs, inputs, locator strings and assertion text out of
|
|
27
|
+
generated regression tests, including unknown query keys and path tokens. Read
|
|
28
|
+
them from environment variables; expose names and step purposes in `required_env`.
|
|
29
|
+
Omit finding prose from file contents and filenames. Preserve base-URL overrides.
|
|
30
|
+
- Add opt-in `save_private_values` for retaining a local, Git-ignored JSON values
|
|
31
|
+
file with owner-only POSIX permissions. Standalone tests can load it through
|
|
32
|
+
`ARGUS_VALUES_FILE`; explicit environment values take precedence. No values file
|
|
33
|
+
is written by default, and values are never included in tool results. Export
|
|
34
|
+
identities reject another test's file or environment before any browser actions.
|
|
35
|
+
- Withhold raw browser and subprocess error output from generated test failures
|
|
36
|
+
and `export_test` results. Keep explicit symptom, incomplete, login-wall and
|
|
37
|
+
missing-variable diagnostics. Missing input values block execution.
|
|
38
|
+
- Stop incomplete exports before browser actions instead of silently skipping
|
|
39
|
+
unsupported steps. `export_test(run=True)` reports `incomplete` without running.
|
|
40
|
+
- Shorten the project overview, move detailed usage and tool profiles to
|
|
41
|
+
`docs/usage.md`, and consolidate public documentation in English.
|
|
42
|
+
|
|
43
|
+
## Claude plugin 0.12.1 - 2026-10-28
|
|
44
|
+
|
|
45
|
+
- Use `uv run --isolated --locked` without launcher environment overrides.
|
|
46
|
+
Each server gets a temporary environment populated from the existing lock;
|
|
47
|
+
cached downloads are reused and normal exit removes the environment. The
|
|
48
|
+
launcher neither modifies the plugin directory nor uses the host environment.
|
|
49
|
+
- Keep the server pinned to Argus 0.12.0 and preserve all dependency hashes.
|
|
50
|
+
This is a Claude plugin update, not a new Python package or Codex plugin release.
|
|
51
|
+
|
|
3
52
|
## 0.12.0 - 2026-10-28
|
|
4
53
|
|
|
5
54
|
- Exported regression tests reject HTTP error pages and login walls instead of
|
|
@@ -2,8 +2,6 @@
|
|
|
2
2
|
|
|
3
3
|
Keep the primary README, code, documentation and release notes in English and
|
|
4
4
|
focused on product usage, implementation, verification and contribution guidance.
|
|
5
|
-
`README.zh-CN.md` is the dedicated translation exception; keep its version,
|
|
6
|
-
commands, capabilities and limitations synchronized with the English README.
|
|
7
5
|
|
|
8
6
|
Use Python 3.11 or newer. CI runs 3.11 and 3.12. Browser checks are local and do
|
|
9
7
|
not call a model provider:
|
|
@@ -30,7 +28,7 @@ separate macOS manual check; Linux CI does not claim to test that integration.
|
|
|
30
28
|
|
|
31
29
|
- Tool arguments, MCP metadata and profile selection: `entrypoints/mcp/`.
|
|
32
30
|
- A new QA operation: the appropriate `application/` service and `OPERATIONS`
|
|
33
|
-
registry; add catalog metadata,
|
|
31
|
+
registry; add catalog metadata, the usage reference and a protocol test.
|
|
34
32
|
- Pure rules and typed data: `domain/`; no browser, MCP or HTTP imports here.
|
|
35
33
|
- New verification: register an async strategy on `runtime.verifiers`. Return
|
|
36
34
|
an explicit receipt with `attempted` and a three-state `reproduced` result.
|
|
@@ -118,10 +116,12 @@ package, so they move only after PyPI has the new version. Raise the
|
|
|
118
116
|
version in both plugin manifests, run `uv lock` in `plugins/claude` and copy its
|
|
119
117
|
`uv.lock` to `plugins/codex`. Then update the versioned `uvx` commands in both
|
|
120
118
|
plugin READMEs and skills, including the locked Playwright version, and run
|
|
121
|
-
`pytest plugins/tests`.
|
|
122
|
-
|
|
123
|
-
|
|
124
|
-
|
|
119
|
+
`pytest plugins/tests`. Keep both launch checks aligned with the published
|
|
120
|
+
runtime's web tool count and required capabilities, including `export_test`.
|
|
121
|
+
For plugin-only fixes, raise only the affected plugin manifest version; keep
|
|
122
|
+
the published runtime dependency and its lock unchanged. Plugin and runtime
|
|
123
|
+
versions may therefore differ. Hosts update installed plugins only when the
|
|
124
|
+
manifest version changes, and the skills' `--doctor` step assumes the web checks added
|
|
125
125
|
in 0.12.
|
|
126
126
|
MCPRush's default listing should expose the 29-tool web profile, not the
|
|
127
127
|
`evaluation` baseline tools; preserve manually declared schemas when refreshing
|
|
@@ -0,0 +1,181 @@
|
|
|
1
|
+
Metadata-Version: 2.5
|
|
2
|
+
Name: argus-testing
|
|
3
|
+
Version: 0.12.2
|
|
4
|
+
Summary: Testing MCP for coding agents: exercise real user journeys, find bugs, and verify them with evidence and persistence checks
|
|
5
|
+
Project-URL: Homepage, https://yichenwu.dev/projects/argus
|
|
6
|
+
Project-URL: Documentation, https://github.com/chriswu727/argus#readme
|
|
7
|
+
Project-URL: Repository, https://github.com/chriswu727/argus
|
|
8
|
+
Project-URL: Issues, https://github.com/chriswu727/argus/issues
|
|
9
|
+
Author: Yichen Wu
|
|
10
|
+
License: MIT
|
|
11
|
+
License-File: LICENSE
|
|
12
|
+
Keywords: ai-agent,autonomous-testing,browser-testing,bug-finding,claude-code,codex,exploratory-testing,mcp,playwright,qa,qa-engineer,software-testing,testing
|
|
13
|
+
Classifier: Development Status :: 3 - Alpha
|
|
14
|
+
Classifier: Intended Audience :: Developers
|
|
15
|
+
Classifier: License :: OSI Approved :: MIT License
|
|
16
|
+
Classifier: Programming Language :: Python :: 3
|
|
17
|
+
Classifier: Topic :: Software Development :: Testing
|
|
18
|
+
Requires-Python: >=3.11
|
|
19
|
+
Requires-Dist: certifi>=2026.10.23
|
|
20
|
+
Requires-Dist: click>=8.0.0
|
|
21
|
+
Requires-Dist: cryptography>=50.0.2
|
|
22
|
+
Requires-Dist: filelock<5,>=4.2.4
|
|
23
|
+
Requires-Dist: mcp[cli]<2.0.0,>=1.30.0
|
|
24
|
+
Requires-Dist: pillow>=13.0.0
|
|
25
|
+
Requires-Dist: playwright>=1.40.0
|
|
26
|
+
Requires-Dist: pydantic>=2.0.0
|
|
27
|
+
Requires-Dist: pyjwt>=2.15.1
|
|
28
|
+
Requires-Dist: rich>=13.0.0
|
|
29
|
+
Requires-Dist: starlette<2,>=0.27
|
|
30
|
+
Requires-Dist: typing-extensions>=4.9
|
|
31
|
+
Requires-Dist: uvicorn<1,>=0.31
|
|
32
|
+
Provides-Extra: dev
|
|
33
|
+
Requires-Dist: httpx<1,>=0.27; extra == 'dev'
|
|
34
|
+
Requires-Dist: mypy<3,>=2.4; extra == 'dev'
|
|
35
|
+
Requires-Dist: pip-audit<3,>=2.9; extra == 'dev'
|
|
36
|
+
Requires-Dist: pytest-asyncio>=0.23.0; extra == 'dev'
|
|
37
|
+
Requires-Dist: pytest>=8.0.0; extra == 'dev'
|
|
38
|
+
Requires-Dist: ruff<0.18,>=0.17; extra == 'dev'
|
|
39
|
+
Provides-Extra: mac
|
|
40
|
+
Requires-Dist: pyobjc-framework-applicationservices>=10.0; (sys_platform == 'darwin') and extra == 'mac'
|
|
41
|
+
Requires-Dist: pyobjc-framework-cocoa>=10.0; (sys_platform == 'darwin') and extra == 'mac'
|
|
42
|
+
Requires-Dist: pyobjc-framework-quartz>=10.0; (sys_platform == 'darwin') and extra == 'mac'
|
|
43
|
+
Description-Content-Type: text/markdown
|
|
44
|
+
|
|
45
|
+
<!-- mcp-name: io.github.chriswu727/argus -->
|
|
46
|
+
|
|
47
|
+
<div align="center">
|
|
48
|
+
|
|
49
|
+
<img src="https://raw.githubusercontent.com/chriswu727/argus/main/assets/argus-icon.png" alt="Argus logo" width="112">
|
|
50
|
+
|
|
51
|
+
# Argus
|
|
52
|
+
|
|
53
|
+
### Browser testing for your coding agent.
|
|
54
|
+
|
|
55
|
+
[](https://pypi.org/project/argus-testing/)
|
|
56
|
+
[](https://github.com/chriswu727/argus/actions/workflows/ci.yml)
|
|
57
|
+
[](https://pypi.org/project/argus-testing/)
|
|
58
|
+
[](LICENSE)
|
|
59
|
+
|
|
60
|
+
[Quick start](#quick-start) · [What you get](#what-you-get) · [Engineering](#engineering) · [Documentation](#documentation) · [Product page](https://yichenwu.dev/projects/argus)
|
|
61
|
+
|
|
62
|
+
</div>
|
|
63
|
+
|
|
64
|
+
Argus is an [MCP](https://modelcontextprotocol.io/) testing server for Claude Code,
|
|
65
|
+
Codex, Cursor and other coding agents. It gives your existing agent the tools to
|
|
66
|
+
test your app the way a testing engineer works through it: explore user journeys,
|
|
67
|
+
check state after each action, try edge cases, verify suspected failures on a fresh
|
|
68
|
+
load and report with evidence.
|
|
69
|
+
The agent decides what to test; Argus handles browser actions, repeatable checks
|
|
70
|
+
and the results.
|
|
71
|
+
|
|
72
|
+
> Test my web app at http://localhost:3000 and look for bugs.
|
|
73
|
+
|
|
74
|
+
Your agent can explore user flows, inspect what changed after each action and
|
|
75
|
+
recheck suspected failures on a fresh page load. You can also give it specific
|
|
76
|
+
goals, such as testing account settings or checking whether saved data survives
|
|
77
|
+
a refresh.
|
|
78
|
+
|
|
79
|
+
## What you get
|
|
80
|
+
|
|
81
|
+
- **A report you can act on.** Findings include expected and actual behavior,
|
|
82
|
+
reproduction steps, screenshots and verification results. Reports also show
|
|
83
|
+
which testing goals were exercised, blocked or left unfinished.
|
|
84
|
+
- **A separate verification step.** Argus rechecks the specified symptom before
|
|
85
|
+
marking it reproduced. Observations and inconclusive checks remain distinct
|
|
86
|
+
from reproduced findings.
|
|
87
|
+
- **A regression test to keep.** Supported findings can be exported as standalone
|
|
88
|
+
Playwright tests. Confirm that the test fails on the bug before adding it to
|
|
89
|
+
your suite; it should pass after the behavior is fixed.
|
|
90
|
+
|
|
91
|
+
<div align="center">
|
|
92
|
+
<img src="https://raw.githubusercontent.com/chriswu727/argus/main/assets/report.png" alt="Example Argus report showing findings and reproduction evidence" width="800">
|
|
93
|
+
</div>
|
|
94
|
+
|
|
95
|
+
For example, a task may disappear after deletion but return after a refresh.
|
|
96
|
+
Argus can check that behavior on fresh loads, retain the evidence and export a
|
|
97
|
+
test for the persistence failure. The receipt confirms that symptom; it does
|
|
98
|
+
not identify the root cause or fix the application.
|
|
99
|
+
|
|
100
|
+
Reports are available as **HTML, JSON, JUnit and SARIF**. Test exports currently
|
|
101
|
+
support findings verified through fresh-load text checks. Uploads, restored
|
|
102
|
+
browser state and other unsupported steps require manual translation.
|
|
103
|
+
|
|
104
|
+
## Quick start
|
|
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**:
|
|
115
|
+
|
|
116
|
+
```bash
|
|
117
|
+
claude mcp add argus -- uvx --from argus-testing argus-mcp
|
|
118
|
+
```
|
|
119
|
+
|
|
120
|
+
Or **Codex**:
|
|
121
|
+
|
|
122
|
+
```bash
|
|
123
|
+
codex mcp add argus -- uvx --from argus-testing argus-mcp
|
|
124
|
+
```
|
|
125
|
+
|
|
126
|
+
Then ask your agent to test your app. Use test accounts and data you can change:
|
|
127
|
+
replaying a journey can repeat its writes.
|
|
128
|
+
|
|
129
|
+
[Cursor, other MCP clients and troubleshooting](docs/usage.md#quick-start) ·
|
|
130
|
+
[Claude plugin](plugins/claude) · [Codex plugin](plugins/codex)
|
|
131
|
+
|
|
132
|
+
**0.12.2**: regression checks preserve complete verification conditions and treat
|
|
133
|
+
intermittent results as inconclusive. Exported tests stop on missing or ambiguous
|
|
134
|
+
row targets. Regenerate older tests and follow the [journal migration steps](docs/usage.md).
|
|
135
|
+
See the [releases](https://github.com/chriswu727/argus/releases) and
|
|
136
|
+
[changelog](CHANGELOG.md#0122---2026-11-08).
|
|
137
|
+
|
|
138
|
+
## Engineering
|
|
139
|
+
|
|
140
|
+
Where to start reading the code:
|
|
141
|
+
|
|
142
|
+
- **Verification with explicit outcomes.** A completed tool call is not a passed
|
|
143
|
+
application test. Checks return evidence and a reproduced, not reproduced or
|
|
144
|
+
inconclusive outcome. HTTP errors and login walls stop exported checks from
|
|
145
|
+
treating a missing symptom as a fix.
|
|
146
|
+
- **Recorded actions become reusable tests.** Exports use the controls resolved
|
|
147
|
+
during the session, preserve iframe and row scope, and check the symptom on a
|
|
148
|
+
fresh load. The [export tests](tests/test_export_test.py) exercise both buggy
|
|
149
|
+
and fixed versions of local fixtures.
|
|
150
|
+
- **Separated runtime and adapters.** Session lifecycle, browser
|
|
151
|
+
operations, verification and reporting are separated. The host agent supplies
|
|
152
|
+
the model; Argus does not run its own model loop or require another API key.
|
|
153
|
+
See the [architecture](docs/architecture.md) and [design decisions](docs/decisions.md).
|
|
154
|
+
|
|
155
|
+
Argus is still an alpha project. A fresh load normally retains authentication
|
|
156
|
+
and browser storage; it is not a reset of the backend. A reproduced symptom is
|
|
157
|
+
not proof of complete coverage or of a requirements violation. Generated tests
|
|
158
|
+
and screenshots should be reviewed before sharing, especially with sensitive
|
|
159
|
+
application data. [Privacy details](PRIVACY.md).
|
|
160
|
+
|
|
161
|
+
## Evaluation
|
|
162
|
+
|
|
163
|
+
The deterministic fixtures cover **34 seeded scenarios**. That measures what
|
|
164
|
+
the tools can exercise with scripted steps, not an agent's bug-discovery rate.
|
|
165
|
+
In a [small paired Codex evaluation](docs/evaluation-0.8.md), both Argus and the
|
|
166
|
+
browser baseline found all 5 predeclared defects. Argus also found one fixture
|
|
167
|
+
defect identified after the run and used more time and tokens. These results
|
|
168
|
+
do not establish general superiority.
|
|
169
|
+
|
|
170
|
+
The [evaluation guide](docs/evaluation.md) documents the method, limits and
|
|
171
|
+
reproduction commands. Browser and unit tests run locally without model calls;
|
|
172
|
+
model-based evaluations are separate and opt-in.
|
|
173
|
+
|
|
174
|
+
## Documentation
|
|
175
|
+
|
|
176
|
+
- [Usage and tool reference](docs/usage.md) — client setup, profiles, configuration and regression workflows.
|
|
177
|
+
- [Agent protocol](docs/agent-protocol.md) — structured results, evidence and coverage tracking.
|
|
178
|
+
- [Contributing](CONTRIBUTING.md) — local checks and implementation boundaries.
|
|
179
|
+
- [Changelog](CHANGELOG.md) · [Security](SECURITY.md) · [Privacy](PRIVACY.md).
|
|
180
|
+
|
|
181
|
+
Built by [Yichen Wu](https://github.com/chriswu727) · [yichenwu.dev](https://yichenwu.dev)
|
|
@@ -32,7 +32,10 @@ credential-bearing URL parameters and headers are redacted in captured requests.
|
|
|
32
32
|
All built-in report formats use a common redacted copy of the report data,
|
|
33
33
|
including visited URLs and nested reproduction evidence. This does not modify
|
|
34
34
|
the live session or the journal's executable verification conditions. Journals
|
|
35
|
-
remain private local artifacts and may retain sensitive verification URLs
|
|
35
|
+
remain private local artifacts and may retain sensitive verification URLs,
|
|
36
|
+
full assertion text and verification preconditions. The complete execution
|
|
37
|
+
predicate stays separate from the public receipt's text preview; a preview
|
|
38
|
+
is not a substitute for reviewing artifacts before sharing them.
|
|
36
39
|
Screenshots and unlabeled application-specific secrets still require manual review.
|
|
37
40
|
|
|
38
41
|
Cookie and storage inspection hides values unless the operator enables
|
|
@@ -40,6 +43,31 @@ Cookie and storage inspection hides values unless the operator enables
|
|
|
40
43
|
capture/restore, which intentionally preserves raw browser state. Local file
|
|
41
44
|
inputs require approved directories in `ARGUS_FILE_ROOTS`; see [Security](SECURITY.md).
|
|
42
45
|
|
|
46
|
+
Since 0.12.1, Argus keeps application URLs, typed/selected values, locator strings and
|
|
47
|
+
assertion text out of generated regression test source. Finding descriptions are
|
|
48
|
+
omitted, and filenames use the run ID and finding index. All application strings
|
|
49
|
+
are supplied through environment variables. The tool returns variable names with
|
|
50
|
+
fixed step/purpose labels, not values; an immediate run passes the values in the
|
|
51
|
+
child environment without writing a values file by default. Only explicit
|
|
52
|
+
`save_private_values=true` retains a raw JSON file in a new local directory;
|
|
53
|
+
POSIX permissions are 0700 for the directory and 0600 for the file. A `.gitignore`
|
|
54
|
+
ignores the directory's contents. This is not encryption and does not protect
|
|
55
|
+
against forced Git adds, archive uploads, backups or access by the local account.
|
|
56
|
+
The file persists until the user deletes it. Never publish it. `ARGUS_VALUES_FILE`
|
|
57
|
+
lets its matching generated test load it without changing the runner's process
|
|
58
|
+
environment. File and environment values require a matching export identity to
|
|
59
|
+
prevent accidentally replaying another test's values.
|
|
60
|
+
Generated exceptions suppress raw
|
|
61
|
+
browser errors; tool results never forward raw subprocess output. These safeguards
|
|
62
|
+
are available in 0.12.1. Earlier 0.12.0 exports parameterize only recognized
|
|
63
|
+
sensitive inputs; regenerate old tests after upgrading.
|
|
64
|
+
|
|
65
|
+
Environment variables are not encrypted or isolated from the local account, child
|
|
66
|
+
processes, debuggers, process dumps or CI administrators. Do not enable runner
|
|
67
|
+
options that dump environments/locals or publish CI secrets, reports or screenshots.
|
|
68
|
+
Review any manual edits to generated tests before sharing them. Export protection
|
|
69
|
+
does not sanitize the separate application evidence described above.
|
|
70
|
+
|
|
43
71
|
## Where data can go
|
|
44
72
|
|
|
45
73
|
- The MCP host and its configured model provider receive the Argus tool inputs
|
|
@@ -0,0 +1,137 @@
|
|
|
1
|
+
<!-- mcp-name: io.github.chriswu727/argus -->
|
|
2
|
+
|
|
3
|
+
<div align="center">
|
|
4
|
+
|
|
5
|
+
<img src="https://raw.githubusercontent.com/chriswu727/argus/main/assets/argus-icon.png" alt="Argus logo" width="112">
|
|
6
|
+
|
|
7
|
+
# Argus
|
|
8
|
+
|
|
9
|
+
### Browser testing for your coding agent.
|
|
10
|
+
|
|
11
|
+
[](https://pypi.org/project/argus-testing/)
|
|
12
|
+
[](https://github.com/chriswu727/argus/actions/workflows/ci.yml)
|
|
13
|
+
[](https://pypi.org/project/argus-testing/)
|
|
14
|
+
[](LICENSE)
|
|
15
|
+
|
|
16
|
+
[Quick start](#quick-start) · [What you get](#what-you-get) · [Engineering](#engineering) · [Documentation](#documentation) · [Product page](https://yichenwu.dev/projects/argus)
|
|
17
|
+
|
|
18
|
+
</div>
|
|
19
|
+
|
|
20
|
+
Argus is an [MCP](https://modelcontextprotocol.io/) testing server for Claude Code,
|
|
21
|
+
Codex, Cursor and other coding agents. It gives your existing agent the tools to
|
|
22
|
+
test your app the way a testing engineer works through it: explore user journeys,
|
|
23
|
+
check state after each action, try edge cases, verify suspected failures on a fresh
|
|
24
|
+
load and report with evidence.
|
|
25
|
+
The agent decides what to test; Argus handles browser actions, repeatable checks
|
|
26
|
+
and the results.
|
|
27
|
+
|
|
28
|
+
> Test my web app at http://localhost:3000 and look for bugs.
|
|
29
|
+
|
|
30
|
+
Your agent can explore user flows, inspect what changed after each action and
|
|
31
|
+
recheck suspected failures on a fresh page load. You can also give it specific
|
|
32
|
+
goals, such as testing account settings or checking whether saved data survives
|
|
33
|
+
a refresh.
|
|
34
|
+
|
|
35
|
+
## What you get
|
|
36
|
+
|
|
37
|
+
- **A report you can act on.** Findings include expected and actual behavior,
|
|
38
|
+
reproduction steps, screenshots and verification results. Reports also show
|
|
39
|
+
which testing goals were exercised, blocked or left unfinished.
|
|
40
|
+
- **A separate verification step.** Argus rechecks the specified symptom before
|
|
41
|
+
marking it reproduced. Observations and inconclusive checks remain distinct
|
|
42
|
+
from reproduced findings.
|
|
43
|
+
- **A regression test to keep.** Supported findings can be exported as standalone
|
|
44
|
+
Playwright tests. Confirm that the test fails on the bug before adding it to
|
|
45
|
+
your suite; it should pass after the behavior is fixed.
|
|
46
|
+
|
|
47
|
+
<div align="center">
|
|
48
|
+
<img src="https://raw.githubusercontent.com/chriswu727/argus/main/assets/report.png" alt="Example Argus report showing findings and reproduction evidence" width="800">
|
|
49
|
+
</div>
|
|
50
|
+
|
|
51
|
+
For example, a task may disappear after deletion but return after a refresh.
|
|
52
|
+
Argus can check that behavior on fresh loads, retain the evidence and export a
|
|
53
|
+
test for the persistence failure. The receipt confirms that symptom; it does
|
|
54
|
+
not identify the root cause or fix the application.
|
|
55
|
+
|
|
56
|
+
Reports are available as **HTML, JSON, JUnit and SARIF**. Test exports currently
|
|
57
|
+
support findings verified through fresh-load text checks. Uploads, restored
|
|
58
|
+
browser state and other unsupported steps require manual translation.
|
|
59
|
+
|
|
60
|
+
## Quick start
|
|
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**:
|
|
71
|
+
|
|
72
|
+
```bash
|
|
73
|
+
claude mcp add argus -- uvx --from argus-testing argus-mcp
|
|
74
|
+
```
|
|
75
|
+
|
|
76
|
+
Or **Codex**:
|
|
77
|
+
|
|
78
|
+
```bash
|
|
79
|
+
codex mcp add argus -- uvx --from argus-testing argus-mcp
|
|
80
|
+
```
|
|
81
|
+
|
|
82
|
+
Then ask your agent to test your app. Use test accounts and data you can change:
|
|
83
|
+
replaying a journey can repeat its writes.
|
|
84
|
+
|
|
85
|
+
[Cursor, other MCP clients and troubleshooting](docs/usage.md#quick-start) ·
|
|
86
|
+
[Claude plugin](plugins/claude) · [Codex plugin](plugins/codex)
|
|
87
|
+
|
|
88
|
+
**0.12.2**: regression checks preserve complete verification conditions and treat
|
|
89
|
+
intermittent results as inconclusive. Exported tests stop on missing or ambiguous
|
|
90
|
+
row targets. Regenerate older tests and follow the [journal migration steps](docs/usage.md).
|
|
91
|
+
See the [releases](https://github.com/chriswu727/argus/releases) and
|
|
92
|
+
[changelog](CHANGELOG.md#0122---2026-11-08).
|
|
93
|
+
|
|
94
|
+
## Engineering
|
|
95
|
+
|
|
96
|
+
Where to start reading the code:
|
|
97
|
+
|
|
98
|
+
- **Verification with explicit outcomes.** A completed tool call is not a passed
|
|
99
|
+
application test. Checks return evidence and a reproduced, not reproduced or
|
|
100
|
+
inconclusive outcome. HTTP errors and login walls stop exported checks from
|
|
101
|
+
treating a missing symptom as a fix.
|
|
102
|
+
- **Recorded actions become reusable tests.** Exports use the controls resolved
|
|
103
|
+
during the session, preserve iframe and row scope, and check the symptom on a
|
|
104
|
+
fresh load. The [export tests](tests/test_export_test.py) exercise both buggy
|
|
105
|
+
and fixed versions of local fixtures.
|
|
106
|
+
- **Separated runtime and adapters.** Session lifecycle, browser
|
|
107
|
+
operations, verification and reporting are separated. The host agent supplies
|
|
108
|
+
the model; Argus does not run its own model loop or require another API key.
|
|
109
|
+
See the [architecture](docs/architecture.md) and [design decisions](docs/decisions.md).
|
|
110
|
+
|
|
111
|
+
Argus is still an alpha project. A fresh load normally retains authentication
|
|
112
|
+
and browser storage; it is not a reset of the backend. A reproduced symptom is
|
|
113
|
+
not proof of complete coverage or of a requirements violation. Generated tests
|
|
114
|
+
and screenshots should be reviewed before sharing, especially with sensitive
|
|
115
|
+
application data. [Privacy details](PRIVACY.md).
|
|
116
|
+
|
|
117
|
+
## Evaluation
|
|
118
|
+
|
|
119
|
+
The deterministic fixtures cover **34 seeded scenarios**. That measures what
|
|
120
|
+
the tools can exercise with scripted steps, not an agent's bug-discovery rate.
|
|
121
|
+
In a [small paired Codex evaluation](docs/evaluation-0.8.md), both Argus and the
|
|
122
|
+
browser baseline found all 5 predeclared defects. Argus also found one fixture
|
|
123
|
+
defect identified after the run and used more time and tokens. These results
|
|
124
|
+
do not establish general superiority.
|
|
125
|
+
|
|
126
|
+
The [evaluation guide](docs/evaluation.md) documents the method, limits and
|
|
127
|
+
reproduction commands. Browser and unit tests run locally without model calls;
|
|
128
|
+
model-based evaluations are separate and opt-in.
|
|
129
|
+
|
|
130
|
+
## Documentation
|
|
131
|
+
|
|
132
|
+
- [Usage and tool reference](docs/usage.md) — client setup, profiles, configuration and regression workflows.
|
|
133
|
+
- [Agent protocol](docs/agent-protocol.md) — structured results, evidence and coverage tracking.
|
|
134
|
+
- [Contributing](CONTRIBUTING.md) — local checks and implementation boundaries.
|
|
135
|
+
- [Changelog](CHANGELOG.md) · [Security](SECURITY.md) · [Privacy](PRIVACY.md).
|
|
136
|
+
|
|
137
|
+
Built by [Yichen Wu](https://github.com/chriswu727) · [yichenwu.dev](https://yichenwu.dev)
|
|
@@ -1,5 +1,7 @@
|
|
|
1
1
|
from __future__ import annotations
|
|
2
2
|
|
|
3
|
+
import hashlib
|
|
4
|
+
import json
|
|
3
5
|
import re
|
|
4
6
|
from pathlib import Path
|
|
5
7
|
from typing import Optional
|
|
@@ -64,15 +66,23 @@ def _journal_path(origin: str) -> Path:
|
|
|
64
66
|
|
|
65
67
|
|
|
66
68
|
def _bug_fingerprint(bug: "Bug") -> str:
|
|
67
|
-
|
|
68
|
-
|
|
69
|
-
|
|
70
|
-
|
|
71
|
-
|
|
72
|
-
|
|
73
|
-
|
|
74
|
-
|
|
75
|
-
|
|
69
|
+
r = bug.verification_spec or bug.reproduction_receipt or {}
|
|
70
|
+
identity = {
|
|
71
|
+
key: r.get(key)
|
|
72
|
+
for key in (
|
|
73
|
+
"version",
|
|
74
|
+
"kind",
|
|
75
|
+
"at_url",
|
|
76
|
+
"expect",
|
|
77
|
+
"target_text",
|
|
78
|
+
"expect_status",
|
|
79
|
+
"clear_storage",
|
|
80
|
+
"login_page",
|
|
81
|
+
)
|
|
82
|
+
}
|
|
83
|
+
identity["at_url"] = identity["at_url"] or bug.url
|
|
84
|
+
payload = json.dumps(identity, sort_keys=True, ensure_ascii=True)
|
|
85
|
+
return f"{bug.type.value}|{hashlib.sha256(payload.encode()).hexdigest()}"
|
|
76
86
|
|
|
77
87
|
|
|
78
88
|
def _journal_entries(origin: str, *, strict: bool = False) -> list:
|
|
@@ -109,6 +119,8 @@ def _write_journal(s: "Session") -> None:
|
|
|
109
119
|
verify["expect_status"] = expected_status
|
|
110
120
|
else:
|
|
111
121
|
verify.update({"expect": expect, "target_text": target})
|
|
122
|
+
if b.verification_spec:
|
|
123
|
+
verify = dict(b.verification_spec)
|
|
112
124
|
fresh.append(
|
|
113
125
|
{
|
|
114
126
|
"fingerprint": _bug_fingerprint(b),
|