CLAUDE.md

CLAUDE.md — WebRunner

Project Overview

WebRunner (je_web_runner) is a cross-platform web automation framework built on Selenium. It supports multi-browser parallel execution, JSON-driven action scripts, report generation, and remote automation via TCP sockets.

• Language: Python 3.10+
• Dependencies: selenium, requests, python-dotenv, webdriver-manager
• Package: je_web_runner (stable) / je_web_runner_dev (dev)

Development Commands

# Install dependencies
pip install -r requirements.txt

# Install dev dependencies
pip install -r dev_requirements.txt

# Run tests
python -m pytest test/

# Build package
python -m build

Project Structure

The tree below lists the main packages. Every utils/ subpackage with a one-line summary, by layer (core engine, je_web_runner.api facade themes, other): docs/reference/utils_index.md, generated by python scripts/gen_utils_index.py; test/unit_test/test_utils_index.py fails when it is stale.

je_web_runner/
├── __init__.py              # Public API exports
├── __main__.py              # CLI entry point
├── element/                 # WebElement interaction (Wrapper pattern)
├── manager/                 # Multi-driver management (Manager pattern)
├── webdriver/               # WebDriver wrapper & options (Facade pattern)
└── utils/
    ├── callback/            # Event-driven callback executor (Observer pattern)
    ├── exception/           # Custom exception hierarchy
    ├── executor/            # Action executor engine (Command pattern)
    ├── generate_report/     # HTML/JSON/XML report generators (Strategy pattern)
    ├── json/                # JSON file operations
    ├── logging/             # Rotating file handler
    ├── package_manager/     # Dynamic package loading (Plugin pattern)
    ├── project/             # Project template generator (Template pattern)
    ├── socket_server/       # TCP socket server for remote control
    ├── test_object/         # Test object & record classes (Value Object pattern)
    ├── test_record/         # Action recording
    ├── xml/                 # XML utilities
    ├── chrome_profile/      # Persistent Chrome profile + stealth + snapshot/sync-back
    ├── failure_triage/      # AI failure root-cause analysis on failure bundles
    ├── flake_detector/      # Time-decayed flake scoring + quarantine registry
    ├── locator_health/      # Project-wide locator audit + upgrade suggestions
    ├── device_cloud/        # Real-device cloud (BrowserStack/Sauce/LambdaTest) connector
    ├── otel_bridge/         # W3C traceparent injection for distributed tracing
    ├── mutation_testing/    # Action JSON mutation testing (kill rate / score)
    ├── otp_interceptor/     # MailHog/Mailpit/IMAP/SMS OTP polling for 2FA flows
    ├── download_verify/     # PDF / CSV / Excel / JSON / SHA256 download assertions
    ├── test_auto_repair/    # LLM-driven test rewrite from failure + git diff
    ├── edge_case_generator/ # LLM edge-case variant generator (complement to mutation_testing)
    ├── openapi_to_e2e/      # OpenAPI/Swagger spec → WR_http_* action JSON
    ├── cross_tab_sync/      # Multi-page BroadcastChannel / storage propagation asserts
    ├── visual_ai/           # aHash/dHash/pHash + SSIM-proxy for canvas/chart diff
    ├── test_scheduler/      # Value-density scheduler under time + cloud budget
    ├── walkthrough_docs/    # AI step-by-step SOP / Confluence doc from recorded runs
    ├── live_dashboard/      # Aggregated web UI: runs + flake + quarantine + locators
    ├── ocr_assert/          # OCR-based text assertion for canvas / WebGL / image content
    ├── email_render/        # Capture outbound mail (MailHog/Mailpit/EML) + multi-viewport screenshots
    ├── backend_log_correlator/ # W3C trace_id → Loki/Elasticsearch/file log fetch into failure bundle
    ├── websocket_assert/    # WebSocket frame recorder + count / payload / pubsub assertions
    ├── console_error_budget/ # JS console / unhandled-rejection budget with ignore patterns
    ├── chaos_hooks/         # Seeded chaos injection (offline / throttle / mid-flow reload)
    ├── pr_risk_score/       # Fuse flake / impact / locator / coverage signals into 0-100 PR risk
    ├── flag_matrix/         # Feature-flag combo matrix with constraints + minimal failing subset
    ├── session_to_test/     # rrweb / generic session events → WR action JSON
    ├── exploratory_ai/      # Agentic exploratory tester (observer/planner protocols + RandomPlanner)
    ├── story_to_actions/    # LLM-driven user story / Figma frame → validated WR action JSON
    ├── db_snapshot/         # Per-test DB savepoint/rollback with pluggable backend
    ├── time_freezer/        # Inject Date/Date.now/performance.now patch via CDP for deterministic time tests
    ├── persona_runner/      # Same suite × N personas (admin/free/enterprise) matrix
    ├── token_leak_detector/ # Scan HAR / logs / responses for leaked JWTs, API keys, session tokens
    ├── consent_audit/       # GDPR/CCPA cookie classification + pre-consent / post-reject violation detection
    ├── pii_in_screenshot/   # OCR + PII regex (Luhn-validated card, SSN, TWID) scanner over screenshots
    ├── pseudo_localization/ # ASCII → look-alike + expansion + brackets; detect hard-coded i18n leaks
    ├── screen_reader_runner/ # Walk a11y tree to simulate NVDA/VoiceOver order + flag a11y violations
    ├── forced_colors_mode/  # dark / reduced-motion / forced-colors / high-contrast matrix verification
    ├── sse_assert/          # Server-Sent Events recorder + count/data/JSON-shape/strict-id assertions
    ├── webrtc_assert/       # PeerConnection state / ICE / track / RTP stats assertions
    ├── view_transitions/    # Instrumentation + duration/CLS/group assertions for View Transitions API
    ├── test_dedup_ai/       # Structural + embedding-based semantic dedupe of action JSON files
    ├── multimodal_qa/       # Send screenshot + question to vision LLM, parse pass/fail/notes envelope
    ├── prompt_drift_monitor/ # Track LLM-feature output drift via embeddings + lexical anchors
    ├── git_bisect_flake/    # Ledger-only or probe-driven bisect to find regression commit
    ├── test_cost_estimator/ # Cloud-minute × rate-card × CO₂ estimate per suite/runner/test
    ├── slack_digest/        # Render Slack Block-Kit / Teams card / plain-text test digest payload
    ├── webtransport_assert/ # HTTP/3 WebTransport datagram + stream frame recorder + assertions
    ├── indexed_db_explorer/ # IndexedDB snapshot harvest + store/key/index/record assertions
    ├── file_system_access/  # Mock showOpenFilePicker/showSaveFilePicker + record writes
    ├── notifications_audit/ # Notification.requestPermission timing + permission/spam policy checks
    ├── mixed_content_audit/ # HTTP-on-HTTPS detection via HAR + console scanner
    ├── clickjacking_audit/  # X-Frame-Options / frame-ancestors header check + iframe probe
    ├── open_redirect_detector/ # Probe ?redirect=/?next= params with attacker-host payloads
    ├── sri_verify/          # Subresource Integrity hash presence + correctness + crossorigin
    ├── coop_coep_audit/     # crossOriginIsolated COOP/COEP + per-resource CORP/CORS check
    ├── inp_tracker/         # Interaction to Next Paint instrumentation + p98 + budget
    ├── hydration_check/     # SSR hydration mismatch detection (DOM diff + console markers)
    ├── bundle_budget/       # Per-asset-kind byte budget from HAR + biggest-assets ranking
    ├── third_party_budget/  # Third-party vendor classification + req/byte/blocking-ms budgets
    ├── long_animation_frame/ # Long Animation Frame API listener + per-script attribution
    ├── grpc_tester/         # gRPC stub call recorder + gRPC-Web framing/trailer helpers
    ├── webhook_receiver/    # Threaded HTTP server for catching app's outbound webhooks
    ├── idempotency_check/   # Run request twice + compare status/body/state/side-effects
    ├── pagination_audit/    # Walk all pages, detect dups/gaps/cursor-loop/sort violations
    ├── failure_narrator/    # LLM natural-language failure summary from failure_bundle
    ├── repro_minimizer/     # Delta-debugging (ddmin) to shrink failing action list to minimum
    ├── locator_hardener/    # Heuristic fragility score + LLM-suggested stable selectors
    ├── test_categorizer/    # Auto-tag tests as smoke / regression / perf / a11y / data / api
    ├── quarantine_age_report/ # Quarantine entries with age + fresh/lingering/stale/abandoned tiers
    ├── test_debt_dashboard/ # Inventory of skip/xfail/TODO/_skip markers with age + CODEOWNERS
    ├── sla_tracker/         # % suites finishing under SLA threshold, weekly/daily bucketing
    ├── bug_repro_stability/ # Repeat probe N times, classify deterministic/flaky/non-reproducible
    ├── test_owners_map/     # CODEOWNERS parser + override layer + unowned-test audit
    ├── popover_assert/      # <dialog>/popover open/close/invoker assertions
    ├── cookie_store_api/    # Async cookieStore API harvest + change-event assertions
    ├── speculation_rules/   # Speculation Rules (prerender/prefetch) verification + no-double-fire
    ├── web_locks/           # Multi-tab Web Locks contention harness + deadlock/serialise assertions
    ├── storage_buckets/     # Storage Buckets API isolation + durability + IDB-isolation checks
    ├── hydration_streaming/ # Streaming SSR per-boundary timing + arrival/interactive assertions
    ├── memory_pressure_emulate/ # CDP memory/CPU pressure emulation profiles + run-under-profile
    ├── third_party_block_test/ # Vendor-by-vendor block-resilience matrix
    ├── bundle_diff_pr/      # PR bundle delta (added/removed/grew) + markdown report + growth gate
    ├── prompt_injection_scanner/ # LLM jailbreak payload library + canary-leak scan
    ├── cors_matrix/         # CORS preflight matrix probe + credentials/origin policy assertions
    ├── oauth_pkce_replay/   # Replay OAuth state/PKCE verifier; confirm server rejects
    ├── cookie_chips_audit/  # CHIPS Partitioned cookie compliance auditor
    ├── sbom_diff/           # CycloneDX SBOM diff (added/removed/upgrade/license/vuln)
    ├── failure_auto_tag/    # Heuristic + LLM failure auto-tagger (flaky-locator/timeout/js-error...)
    ├── test_self_describe/  # Reverse-engineer Gherkin Given/When/Then from action JSON
    ├── pr_title_generator/  # Conventional-Commits PR title from diff + commit history
    ├── action_refactor_suggester/ # Rule-based action-JSON refactor smells (hard sleep / positional xpath...)
    ├── rtl_layout_verify/   # RTL layout direction / logical-property / bidi-isolation audit
    ├── dst_boundary_test/   # DST spring-forward/fall-back gap & overlap detection + scheduled-fire model
    ├── number_currency_locale/ # Number/currency/date locale-format assertion helpers
    ├── wcag22_touch_target/ # WCAG 2.2 SC 2.5.8 target-size auditor with spacing-circle exception
    ├── graphql_n_plus_1/    # N+1 query detector for GraphQL operations
    ├── mq_assert/           # Kafka/RabbitMQ/SQS-style message-queue publish assertions
    ├── grpc_streaming_assert/ # gRPC streaming (unary/server/client/bidi) frame/status/half-close
    ├── webhook_signature_verify/ # GitHub/Stripe/Slack/generic HMAC webhook verifier
    ├── test_roi_scorer/     # Find-rate/cost/coverage/recency-weighted ROI score per test
    ├── pre_merge_gate_dsl/  # Declarative pre-merge gate rules (when/require) over PrFacts
    ├── commit_msg_trigger/  # Parse [skip ci]/[ci e2e]/[ci shard=3/8]/tickets from commit message
    ├── flakiness_graveyard/ # Quarantine/revive/bury ledger with TTL for stale flaky tests
    ├── test_blame_owner/    # CODEOWNERS + git-blame + HEAD + default → test owner chain
    ├── webgpu_pixel_verify/ # WebGPU canvas pixel readback + mean/solid/tile-diff assertions
    ├── webhid_mock/         # WebHID device shim with input/output report harness
    ├── webusb_mock/         # WebUSB device shim with control/bulk transfer capture
    ├── webserial_mock/      # Web Serial UART shim + line write capture
    ├── webcodecs_assert/    # WebCodecs chunk codec/resolution/keyframe/framerate assertions
    ├── speech_api_assert/   # SpeechSynthesis/SpeechRecognition mock + spoke/lang assertions
    ├── webauthn_mock/       # WebAuthn / FIDO2 / Passkey navigator.credentials shim
    ├── credential_management/ # navigator.credentials password/federated autofill mock
    ├── payment_request_assert/ # Payment Request API + Apple/Google Pay sheet validation
    ├── three_d_secure_flow/ # 3DS challenge / frictionless / fallback path assertions
    ├── rag_grounding_assert/ # RAG citation + grounding + hallucination phrase scan
    ├── llm_token_cost_tracker/ # Per-test token/$ ledger + per-model rate card + budget
    ├── streaming_chat_assert/ # TTFT / inter-token gap / UTF-8 / dup-or-OOS chunk assertions
    ├── tool_call_assert/    # LLM tool/function call name+order+arg-schema assertions
    ├── hallucination_probe/ # Ground-truth probe runner + hallucination rate budget
    ├── web_push_assert/     # VAPID subscription + endpoint + userVisibleOnly + showNotification
    ├── background_sync_assert/ # Background Sync register/fire/retry/lastChance assertions
    ├── wake_lock_assert/    # Screen wake lock acquire/release/leak/re-acquire detection
    ├── pip_assert/          # Picture-in-Picture (video + document) enter/exit/size assertions
    ├── web_share_assert/    # navigator.share recorder + payload + fallback assertions
    ├── compression_streams/ # CompressionStream gzip/deflate/brotli round-trip + ratio budget
    ├── compute_pressure/    # Compute Pressure API fake observer + throttle reaction assertions
    ├── touch_gesture/       # tap/swipe/pinch/long-press CDP-frame builder + event assertions
    ├── viewport_audit/      # viewport meta + safe-area-inset + WCAG 1.4.4 scalable audit
    ├── virtual_keyboard/    # visualViewport before/after + keyboard inset + focused-visible
    ├── pull_to_refresh/     # overscroll-behavior + threshold + refresh handler + network refetch
    ├── email_deliverability/ # SPF/DKIM/DMARC + List-Unsubscribe + BCC leak header audit
    ├── inbox_render_outlook/ # Outlook/Gmail/Apple Mail render compatibility pre-flight
    ├── push_delivery/       # FCM/APNs payload size + required fields + PII + collapse + TTL
    ├── lcp_image_audit/     # LCP image preload + no-lazy + fetchpriority=high assertions
    ├── font_loading_strategy/ # @font-face display + size-adjust + FOIT/FOUT/FOFT verification
    ├── resource_hints_audit/ # preload/prefetch/preconnect used vs declared + preload-as audit
    ├── critical_css_audit/  # Inline-CSS-in-head + first-packet budget + preload-blocking-CSS
    ├── lighthouse_regression/ # Lighthouse score regression vs baseline + CWV metric budget
    ├── dom_xss_taint/       # source→sink JS instrumentation + canary-based taint detection
    ├── csp_violation_parser/ # CSP report-uri / report-to parser + recon heuristic
    ├── hsts_preload_audit/  # HSTS preload-list compliance (max-age + includeSubDomains + preload)
    ├── tls_cipher_audit/    # Live TLS handshake + version + cipher allowlist + subject check
    ├── cookie_scope_abuse/  # Session-like cookie scope / HttpOnly / Secure / SameSite audit
    ├── test_dup_dry/        # Structural action-JSON duplicate + prefix-overlap detection
    ├── snapshot_diff_approval/ # Baseline/pending/rejected snapshot register + approval workflow
    ├── failure_cluster_dbscan/ # Failure-message tokeniser + DBSCAN root-cause clustering
    ├── test_naming_lint/    # should_when / given_when_then / camel_subject naming linter
    ├── openapi_drift/       # Live API vs spec drift (undocumented / zombie / status / method)
    ├── api_version_compat/  # Old-client vs new-server backward-compat response/request matrix
    ├── rate_limit_assert/   # 429 / Retry-After / X-RateLimit headers + recovery assertions
    └── har_to_openapi/      # HAR → OpenAPI 3.1 path/method/query/schema reverse engineering

Design Patterns & Architecture

• Facade: WebDriverWrapper abstracts Selenium's complex API into a simplified interface
• Command: Action executor maps string commands to callable functions, enabling JSON-driven automation
• Manager: WebdriverManager coordinates multiple browser instances for parallel execution
• Observer/Callback: callback_executor provides event-driven hooks on action completion
• Strategy: Report generators (HTML/JSON/XML) share a common interface with interchangeable output formats
• Plugin: package_manager dynamically loads external packages into the executor at runtime
• Value Object: TestObject encapsulates immutable locator information (name + strategy)

Coding Standards

Software Engineering Principles

• SOLID: Each module has a single responsibility; depend on abstractions (wrappers), not Selenium internals directly
• DRY: Reuse existing wrappers and utilities; never duplicate element-finding or driver-management logic
• KISS: Prefer clear, readable code over clever abstractions; no premature optimization
• YAGNI: Only implement features that are currently needed; no speculative code

Performance Best Practices

• Use implicit/explicit waits instead of time.sleep() for element synchronization
• Prefer find_element with specific locators (ID, CSS selector) over slow strategies (XPath with text matching)
• Reuse WebDriver instances via the manager instead of creating new ones
• Minimize JavaScript execution calls; batch operations where possible
• Use connection pooling for socket server communications
• Avoid loading unnecessary browser extensions or capabilities

Security Requirements

• Input validation: Validate and sanitize ALL external inputs — URLs, JSON action files, socket messages, CLI arguments
• No arbitrary code execution: Action executor must only call registered commands; never use eval() or exec() on user input
• Socket server: Bind to localhost by default; require explicit configuration for network exposure
• File operations: Use safe path handling; prevent path traversal attacks in file-based action execution
• Credentials: Never log, store, or transmit credentials in plaintext; use python-dotenv for environment-based secrets
• Dependencies: Keep all dependencies up to date; audit for known vulnerabilities regularly
• XSS prevention: Escape all dynamic content in generated HTML reports
• Injection prevention: Parameterize any dynamic values passed to JavaScript execution

Code Quality

• Remove dead code, unused imports, and commented-out blocks — do not leave them in the codebase
• Every function should have a clear purpose; if it's unused, delete it
• Keep modules focused; extract new modules only when a clear boundary exists
• Type hints on all public API functions
• Logging at WARNING+ level via the rotating file handler (WEBRunner.log)

Static Analysis Compliance (SonarQube & Codacy)

All code MUST pass SonarQube and Codacy static analysis without introducing new issues. Follow these rules proactively:

Complexity & Maintainability

• Cognitive complexity ≤ 15 per function (SonarQube python:S3776); extract helpers when nesting grows
• Cyclomatic complexity ≤ 10 per function; split branchy logic into smaller units
• Function length ≤ 75 lines; file length ≤ 750 lines; parameters ≤ 7 per function (python:S107)
• Max nesting depth = 4 (python:S134); use early returns / guard clauses to flatten
• No duplicated code blocks ≥ 3 lines (common-py:DuplicatedBlocks); extract shared logic
• No dead stores — never assign a value that is immediately overwritten or unused (python:S1854)

Naming & Style (PEP 8 enforced)

• Modules / functions / variables: snake_case; classes: PascalCase; constants: UPPER_SNAKE_CASE (python:S116, python:S117)
• No single-letter names except loop indices i, j, k and comprehension vars
• Line length ≤ 120 chars; 4-space indentation; no tabs
• No wildcard imports (from x import *) — python:S2208
• One statement per line; no semicolons

Bug Prevention

• Never use mutable default arguments (def f(x=[])) — use None and initialize inside (python:S5797)
• Never use bare except: — catch specific exceptions (python:S5754); never except Exception without re-raise/log
• Never silently swallow exceptions — pass inside except is forbidden unless justified by a # noqa comment (python:S2486)
• Always close resources — use with for files, sockets, drivers (python:S5042)
• No == comparison with None, True, False — use is / is not (python:S5727)
• No identical expressions on both sides of ==, !=, and, or (python:S1764)
• Self-assignment forbidden (x = x) — python:S1656
• No unreachable code after return / raise / break / continue (python:S1763)

Security (SonarQube hotspots / Codacy bandit)

• Never use eval, exec, compile, __import__ on untrusted input (python:S1523)
• Never use pickle, marshal, shelve to deserialize untrusted data (python:S5135)
• Never use subprocess with shell=True on user input (python:S4721)
• Never use assert for security checks — assertions are stripped with -O (python:S5915)
• Never hard-code credentials, tokens, IPs, or URLs with secrets (python:S2068, python:S1313)
• No insecure hash algorithms (MD5, SHA-1) for security purposes — use SHA-256+ (python:S4790)
• No insecure TLS/SSL — never disable certificate verification (verify=False) — python:S4830
• No predictable random for security tokens — use secrets, not random (python:S2245)
• XML parsing must disable entity expansion to prevent XXE — use defusedxml (python:S2755)

Documentation & Typing

• Public modules, classes, and functions require docstrings (Codacy pylint:missing-docstring)
• Type hints on every public function signature; prefer from __future__ import annotations for forward refs
• No # TODO / # FIXME without an associated issue link (python:S1135)

Testing Quality

• Test functions must contain at least one assertion (python:S2699)
• Never use assert True / assert 1 == 1 as placeholders (python:S2187)
• Disabled tests (@pytest.mark.skip without reason) are flagged — always provide reason="..."

Pre-Commit Verification

Before committing, run the following checks locally and ensure they pass cleanly:

# Lint with project tools (add as needed)
python -m pylint je_web_runner/
python -m flake8 je_web_runner/ --max-line-length=120 --max-complexity=10
python -m bandit -r je_web_runner/ -ll

# Type check
python -m mypy je_web_runner/

# Tests
python -m pytest test/

If SonarQube or Codacy is wired into CI, fix any new issues in the same PR — do not defer them with suppressions unless a justified # noqa: <rule> comment explains why.

Documentation

• README parity. This repository ships README.md (English) alongside the translated README/README_zh-CN.md and README/README_zh-TW.md. All three must stay current with the code.
• When a change alters anything user-facing — features, commands, CLI flags, install/setup steps, configuration, or requirements — update README.md, both translated READMEs (README/README_zh-CN.md and README/README_zh-TW.md), and the affected docs/ sources, all in the same commit, keeping their section structure and content aligned across languages. The translations must reflect the English content, not merely match its headings.
• Never update one language, or README.md alone, and leave the other READMEs or the docs stale. No automated README-parity guard exists, so verify this by hand before committing.

Stage commits, progress.md, docs/updates/ and architecture.md

Workspace rule shared by every repository under D:\Codes (full text: D:\Codes\CLAUDE.md).

• Commit at every stage. A stage is the smallest piece of work that leaves the repository consistent and passes this project's checks (definition of done, tests, lint): one finished progress.md item, or one self-contained step of a larger one. Commit it before starting the next stage, before switching to another repository, and before the session ends. Do not leave work uncommitted across sessions; if a stage cannot be finished, commit the consistent part and record the rest in progress.md.
  • Stage only the files that stage touched (git add <path>, never git add -A), follow this file's commit-message rules, and never add AI attribution.
  • Committing is not pushing: push or open a PR only as this project's branch flow says or when asked.
  • Commit and push frequently. After each big feature — a self-contained stage that passes this project's checks — commit and push to the remote; do not pile up a large batch of work before committing or pushing. Smaller batches collide less with other sessions, let CI catch problems earlier, and are easier to revert. Follow this project's normal branch flow (usually dev).
  • SonarCloud / Codacy findings. When a PR or commit fails a SonarCloud or Codacy check, look the findings up through their APIs instead of guessing. The keys are in environment variables: SonarCloudToken (SonarCloud, e.g. curl -s -u "$SonarCloudToken:" "https://sonarcloud.io/api/issues/search?componentKeys=<key>&pullRequest=<n>&resolved=false") and CODACY_PROJECT_TOKEN (a Codacy project token, valid only for its own project: any other repository answers "Bad credentials", so for a public repository query https://app.codacy.com/api/v3/analysis/organizations/gh/<org>/repositories/<repo>/pull-requests/<n>/issues?status=new without a key). Never reveal a key or any personal credential while doing so: refer to the variables by name only, never echo or print their values, and never put them in files, commit messages, PR or issue text, logs, or any output that leaves the machine.
• progress.md (repository root, tracked) holds outstanding work only: no finished items, no history, no rules.
• docs/updates/ records finished work: one batch file per month (YYYY-MM.md), one entry per piece of work headed ## U-YYYYMMDD-NN · date · title · #tags, and an index with query commands in docs/updates/README.md. When a progress.md item is done, delete it and add a #done entry plus its index row in the same commit.
• architecture.md (repository root) is the short architecture overview: layers, entry points, main flows, extension points, cross-project boundaries. Update it in the same commit whenever a change alters any of those.
• Cross-project contracts are listed in architecture.md §6: what other repositories rely on here (CLI flags, import paths, constructor arguments, file layouts) and what this repository relies on elsewhere. No test here protects them, so never rename or remove one without changing its consumers in the same round, and update §6 whenever a contract is added or changes.

Git & Commit Conventions

• Branch model: main (stable) / dev (development)
• PRs go from dev to main
• Both branches publish to PyPI from CI: a push to main releases je_web_runner (publish_stable.yml), and a push to dev that passes the tests and changes what the package ships releases je_web_runner_dev (the publish-dev job of test_dev.yml, scripts/dev_release.py). Never bump a version by hand; the version in dev.toml is only a floor
• Commit messages: concise, imperative mood (e.g., "Add element validation", "Fix driver cleanup on timeout")
• Do NOT mention any AI tools, assistants, or language models in commit messages — commits must read as standard developer-authored messages
• Do NOT include Co-Authored-By lines referencing AI in commits

Testing

• Tests are located in the test/ directory
• Test files follow the pattern test_*.py
• Use pytest as the test runner
• Integration tests require a browser driver to be available