Who & why
Keeps defects from reaching users — checks the product by eye and hand before release. Thinks like a user (incl. a "mean" one) and a skeptic: not "does the happy path work" but "what can I break". Catches what code can't show: friction, weird edge behavior, illogical flows. Without one: bugs reach users, edges untested, regressions in prod, "works on my machine" becomes the bar.
A day in the life
Morning: parse the story & acceptance criteria, design tests (happy + edges + errors), build a checklist. Day: test by hand, actively break things (empty/huge/double-click/no network/bad data), compare to criteria, file reproducible bug reports. Evening: run regression, recheck fixes, update checklists.
Key skills
Hard: test design (equivalence, boundaries, decision tables), testing techniques, acceptance criteria, defect reporting, regression, basic SQL/DevTools/logs. Soft: skepticism & curiosity, user mindset, attention to detail, persistence, clear defect communication.
Artifacts
Test plan, test cases, bug reports, checklists — all templated in the Artifacts group. Checks against the User Story; release-readiness = DoD.
How AI / vibe-coding boosts the role
Test cases from a story; edge cases; bug report from a symptom; regression checklist; "what else" hints — with ready prompts.
Growth: Junior → Middle → Senior → Lead
Junior: tests to ready cases. Middle: own test design, complex edges, starts automating. Senior: testing strategy & deep product expertise, often grows into SDET. Lead → QA Lead.
Common mistakes
Happy path only; irreproducible bug reports; severity = priority; skipping regression; "tested" without criteria.
What to learn
Test-design techniques, the testing pyramid, severity/priority, regression, checklists, bug management. Read: Lessons Learned in Software Testing; ISTQB Foundation.
Salary (RU)
Junior ~60–110k₽/mo, Middle ~110–190k, Senior ~190–300k. Manual usually below automation; varies — check current data.
Laskoff agent mapping
No direct mapsTo; diff verification is done by the quest-verifier agent (adversarial: happy + edges + errors + neighbor regression), plus a mandatory Playwright browser pass before close.
🤖 Persona prompt
You are a thorough QA engineer who thinks like a user and a skeptic. Help me find defects before users do. For any feature, design tests: the happy path plus mandatory edges, negatives and errors ("what if empty/huge/no network/double-click"). Verify against acceptance criteria, not "seems to work". Write reproducible bug reports: steps, expected vs actual, severity and priority separately, environment, evidence. Remind me about regression (did neighbors break) and flag non-obvious areas (security, accessibility, mobile, locales).