e2e-pipeline

Autonomous UI element mapper. Explores web pages using agent-browser CLI and generates/updates YAML mapping files for E2E testing. Returns page inventory with elements, selectors, and unexplored areas. <example> Context: The e2e-map skill has completed codebase analysis and needs to explore a web app's UI for the first time. user: "Map UI elements:\n base_url: http://localhost:3000\n app: my-app\n auth_profile: ~/.agent-browser/my-app/\n routes: [/dashboard, /users, /settings]\n report_dir: /home/user/project/e2e-reports/20260309-150000-map\n auth_config: {type: manual, verification: {url_not_contains: '/login'}}" assistant: "Reads reference files, runs pre-flight checks, opens browser with auth profile, navigates to each route, takes snapshots and annotated screenshots, extracts interactive elements with stable selectors, generates v2 mapping YAML at .claude/e2e/mappings/my-app.yaml, identifies unexplored dialogs/modals, and returns structured summary." <commentary> The e2e-map skill dispatches this agent after gathering routes from codebase analysis. The agent receives all paths as absolute values and explores autonomously. No existing mapping means full exploration mode. </commentary> </example> <example> Context: The e2e-map skill is re-mapping a single page to update stale selectors. user: "Map UI elements:\n base_url: http://localhost:3000\n app: my-app\n auth_profile: ~/.agent-browser/my-app/\n existing_mapping_path: /home/user/project/.claude/e2e/mappings/my-app.yaml\n target_page: dashboard\n report_dir: /home/user/project/e2e-reports/20260309-160000-map\n auth_config: {type: none}" assistant: "Reads existing mapping, navigates only to the dashboard page, re-snapshots and extracts elements, merges updates into the existing mapping (preserving all other pages), and returns summary with changes." <commentary> When target_page is specified, the agent updates only that page's section in the existing mapping. This is the incremental update path — other pages are preserved byte-for-byte. </commentary> </example> <example> Context: The e2e-map skill is mapping a new app with no prior codebase analysis (browser-only discovery). user: "Map UI elements:\n base_url: http://localhost:5173\n app: admin-panel\n auth_profile: ~/.agent-browser/admin-panel/\n routes: []\n report_dir: /home/user/project/e2e-reports/20260309-170000-map" assistant: "No routes provided — enters discovery mode. Snapshots landing page, extracts navigation links from sidebar/tab bar, builds route list from discovered nav elements, explores each route one level deep, generates mapping YAML, and returns summary with discovered pages." <commentary> Empty routes list triggers discovery mode. The agent also detects auth status since no auth_config was provided — it checks for login redirects and tentatively sets auth.type based on what it finds. </commentary> </example>

Autonomous E2E flow executor. Runs browser-based test flows against web apps using agent-browser CLI. Returns structured pass/fail results with screenshots and trace recording. Operates in isolated subagent context to prevent browser data from polluting main conversation. <example> Context: The e2e-test skill has resolved a single flow YAML and its mapping, and needs to execute the test. user: "Execute E2E flow:\n flow_path: /home/user/project/.claude/e2e/flows/smoke-navigation.yaml\n mapping_path: /home/user/project/.claude/e2e/mappings/my-app.yaml\n auth_profile: ~/.agent-browser/my-app/\n base_url: http://localhost:3000\n app: my-app\n report_dir: /home/user/project/e2e-reports/20260309-143000\n headed: true" assistant: "Reads reference files and flow YAML, runs pre-flight checks, opens browser with auth profile, executes each step sequentially (snapshot -> interact -> validate expectations), captures screenshots on failure, writes report.md, and returns structured summary with pass/fail counts." <commentary> The e2e-test skill dispatches this agent for each flow+mapping pair. The agent receives all paths as absolute values and executes autonomously without needing to resolve flows or mappings. </commentary> </example> <example> Context: The e2e-test skill is running a multi-site suite and dispatches this agent for one site with session isolation enabled. user: "Execute E2E flow:\n flow_path: /home/user/project/.claude/e2e/flows/smoke-navigation.yaml\n mapping_path: /home/user/project/.claude/e2e/mappings/admin-panel.yaml\n auth_profile: ~/.agent-browser/admin-panel/\n base_url: http://localhost:5173\n app: admin-panel\n report_dir: /home/user/project/e2e-reports/20260309-143000/admin-panel\n headed: true\n suite_context: true" assistant: "Detects suite_context=true, uses --session admin-panel on all agent-browser commands for session isolation. Executes flow steps, writes per-site report, returns structured summary." <commentary> When suite_context is true, the agent uses --session to keep browser sessions separate across sites in a multi-site suite. This prevents cookie/localStorage leakage between different apps under test. </commentary> </example>

Parses agent-browser trace.zip files to extract API failures and console errors. Produces a concise trace-analysis.md summary. Runs in isolated context to keep verbose HAR/trace data out of main conversation. <example> Context: The e2e-walkthrough skill completed a walkthrough and needs trace analysis. user: "Analyze trace:\n trace_path: /home/user/project/e2e-reports/20260309-150000/trace.zip\n report_dir: /home/user/project/e2e-reports/20260309-150000" assistant: "[Extracts trace.zip, parses trace.network for HTTP 4xx/5xx failures, parses trace.trace for console errors, filters noise, writes trace-analysis.md, returns structured summary with counts]" <commentary> Agent receives absolute paths, extracts to temp dir, processes all trace files, writes analysis, cleans up temp dir. </commentary> </example> <example> Context: The e2e-test skill finished a flow execution and needs trace analysis with custom noise filtering. user: "Analyze trace:\n trace_path: /tmp/e2e-run/trace.zip\n report_dir: /tmp/e2e-run\n noise_patterns: [\"analytics.google.com\", \"sentry.io\"]" assistant: "[Extracts, parses, applies default + custom noise filters, writes trace-analysis.md showing 2 API failures and 0 console errors]" <commentary> Custom noise_patterns are merged with defaults. Noise is filtered before counting. </commentary> </example> <example> Context: Trace zip is missing trace.network (no network recording was active). user: "Analyze trace:\n trace_path: /home/user/project/e2e-reports/run-1/trace.zip\n report_dir: /home/user/project/e2e-reports/run-1" assistant: "[Extracts zip, finds no trace.network file, reports 0 API failures, parses trace.trace for console errors, writes analysis, returns clean: true]" <commentary> Agent gracefully handles missing files — reports 0 for that category instead of failing. </commentary> </example>

Use when the user explicitly invokes /e2e-dispatch or asks for the E2E operations menu without specifying a particular operation. This is a unified router — for specific operations, prefer e2e-test, e2e-map, e2e-walkthrough, or e2e-skill-ops directly.

Use when creating or updating UI element mappings for E2E testing — mapping page selectors, capturing app structure, or refreshing stale mappings. Triggers on "e2e map", "map the UI", "update mapping", "refresh mapping", "map selectors", "create mapping".

Use when debugging e2e-test/e2e-map/e2e-walkthrough failures, maintaining or adding features to these skills, or evaluating execution results against skill expectations. Triggers on "e2e skill broke", "fix e2e skill", "update e2e skill", "add feature to e2e", "e2e gap analysis", "skill drift", "e2e pipeline issue".

Use when running browser E2E tests against web apps — executing flow files, running smoke suites, testing specific features, recording test runs with video, or batch-testing across sites. Triggers on "run e2e", "browser test", "test the flow", "smoke test", "e2e test", "playwright test", "run the suite", "test this feature in browser", "record test run", "e2e video", "test with video".

Use when walking through UI flows interactively, exploring features with human guidance, or visually validating changes in a browser. Records video by default for evidence and communication. Triggers on "walkthrough", "explore this feature", "walk the UI", "step through", "browse the app", "guided e2e", "interactive walkthrough", "record walkthrough", "video walkthrough".