hathor

Run AND keep healthy hathor's Playwright e2e suite — both mocked and against a LIVE local Sobek backend (E2E_BACKEND=true) — using the exact same sequential flow so the two modes stay identical, AND keeping the suite slim and true (no duplicate coverage, no unit-test-shaped specs). Use this whenever the user wants to run e2e against a real database / local backend, set up the manual-login + JWT-in-session (storageState) handoff, debug the "Loading data…" organisation-selection stall or a 401 from Sobek :37999 in tests, make mocked and live e2e behave the same way, audit that every spec follows the canonical login → select-org → navigate+count → create → edit-verify → edit sequence, or prune the suite (flag pure duplicate coverage and e2e tests that belong in vitest). Trigger eagerly on E2E_BACKEND, `e2e:local-backend`, "live backend e2e", "run e2e against the real DB", organisation-select stalls in tests, "why do my e2e tests hang / 401", "is this spec a duplicate", or "should this be a unit test" — even when

Generate a browsable top-down mermaid flowchart for every Playwright e2e spec by reading each spec's file-level JSDoc, then assemble them into one tmp index.html you can open in a browser. Use this whenever the user wants to visualize, diagram, map out, or "see" the e2e test suite / spec flows — e.g. "draw flowcharts for the e2e specs", "visualize the test workflows", "make a diagram of what each spec does", "render the spec JSDoc as charts", or any request to turn the e2e-tests/**/*.spec.ts keynote JSDoc into mermaid diagrams. Fans the work out across cheap (haiku) subagents — one per spec — so it stays fast and low-cost even on a large suite.