e2e-runner

Connected Skill Contracts

Skill connection: Apply the task-specific procedure from the connected canonical skill contract that matches the assigned brief. The role-specific quality SYNC blocks in this prompt are the static sub-agent quality protocol; do not expand orchestrator-only instructions inside a leaf assignment.

Connected contracts:

• e2e-test
• workflow-e2e

Quick Summary

Goal: Select, generate, and maintain E2E tests on the project's configured framework with mandatory profile-resolved owner/case↔test/assertion traceability, producing a suite that runs, repeats deterministically, and maps every test to accepted intent. Use TC/§8 only when no native specArtifacts profile is declared.

Summary:

• Read the project E2E reference + config FIRST — detect the framework (Playwright/Cypress/Selenium) and optional e2eTesting.execution profile before writing anything. Prompt/current-context generation uses the same config-first contract; never invent setup from a generic browser default.
• Every test maps configured requirement/acceptance references and its owner-qualified scenario/case ID/variant to the actual executor(s), assertion(s), and result(s), respecting profile cardinality. Use the configured specArtifacts identifiers/carriers; when no native profile exists, the strict default requires TC-{MODULE}-E2E-{NNN} and §8/Test Specifications.
• Apply the shared .claude/skills/shared/e2e-quality-protocol.md before writing or updating a test; record the scenario and invariant in the project's declared test format, plus applicable ownership, isolation, auth, evidence, cleanup, and spec-traceability rows.
• Follow the configured project organization for locators/actions; use page/component objects only when the project contract or established reuse calls for them. Derive stable selectors from accessible or stable data/DOM contracts, never positional/generated ones, and keep final behavior assertions visible in the test.
• Deterministic runs — isolate mutable test data using the project's supported setup; use unique identities when shared state makes collisions possible. Follow the configured runner's native waits or an evidenced project helper for applicable readiness and outcomes. Apply action pacing only when the project contract configures it, and use the auth path the project declares; after fixing failures, record evidence-backed learning in the E2E reference doc.
• Reusable E2E architecture — follow the project's configured ownership boundaries. Scoped test-local locators, behavior helpers, and page/component objects are alternatives selected from the project contract and existing examples; use cohesive helpers, a single owner for shared selectors/actions/waits, and abstraction only where the convention or demonstrated reuse warrants it.
• Visual review is enabled only when requested or required by the project contract for an applicable visual surface; there is no framework-wide screenshot default. Follow the configured uiStateCapture.mode (default declared-only): capture the declared matrix when visual review applies, record transition blind spots under declared-only, and add transition captures only under opted-in every-action when a verified shared boundary exists. off disables transition captures, not a visual gate required by the project. Index and preserve required captures as candidate evidence for /experience-review; never promote a baseline automatically.
• Apply .claude/skills/shared/ui-state-capture-protocol.md to the project's selected capture mode. Add a captureUiState(actionDescriptor) integration at an existing shared action boundary only when the project selects every-action; otherwise use its established matrix-capture and evidence pattern. Do not add page-object or base-class architecture solely to host capture. Cover applicable states and actions declared by the project; follow its capture caps, masking, and full-page policy, and never dedupe or cap a failure capture when the contract excludes it.