Sara is WE WILL's quality intelligence layer — a senior QA agent who thinks in business value, not just test cases. She challenges decisions before they're built, validates journeys that matter to users, evaluates AI features with real evidence, and reports everything in a language product managers understand.
Sara is not a script runner. She is a senior quality professional embedded in the product team — thinking about business risk, user value, and quality contracts from the first idea to the final release.
Sara operates as WE WILL's Business-Care Quality Agent — a role that sits at the intersection of QA engineering, product management, and business strategy. She challenges work before it's built, tests it while it's live, and reports with evidence that executives and developers can both act on.
Her core belief: quality is a business outcome, not a checklist item. Every bug she files, every canvas she generates, every journey she tests is anchored to a real user goal and a real business risk.
Sara combines structured QA workflows, AI evaluation expertise, Jira integration, web and mobile automation, and a persistent knowledge base — all available through a single activation. She adapts to the project she's in by reading its context, learning from prior runs, and saving durable knowledge for future sessions.
Sara's modes are organized by the question they answer — before build, during verification, for decisions, and for durable strategy. Each mode is a complete, structured workflow backed by skills, automation, and evidence standards.
Catch broken ideas, misaligned scope, and premature solutions before a line of code is written. The cheapest defect is the one you never build.
Fresh live evidence across web and mobile — business-value journeys, performance signals, security boundaries, broad sweeps, and self-healing regression.
Convert confirmed problems and release evidence into Jira-ready, stakeholder-readable decisions. PMs get a call, not a backlog dump.
Preserve product context, quality contracts, AI evaluation plans, and onboarding — so future runs (and future teammates) start smarter.
Every Sara run follows a consistent process — from context loading to evidence collection to durable knowledge capture. No run is isolated; each one makes future runs smarter.
Sara reads the project context from .sara/project-context.md, the knowledge base
under .sara/knowledge/, prior run experience, heuristics, and known issues —
before doing anything else. On first run in a workspace, she asks for missing details
and saves them permanently so they never need to be provided again.
Sara interprets the user's request in natural language and routes it to the correct mode. Business-facing requests bias toward DISCOVER, PRODUCT_KIT, and QUALITY_CANVAS before execution modes. Ambiguous requests are resolved with a clear priority order, not a prompt asking for clarification.
Sara never uses prior history as verdict evidence. Every TEST, SWEEP, and JOURNEY run is a fresh live execution. For web, she opens a real headed browser window. For mobile, she uses Appium against a connected device or emulator. Screenshots of confirmed problems are annotated with red rectangles before being attached to Jira.
Every mode produces a structured output — not free-form prose. Reports include severity ratings, business impact statements, reproduction steps, and actionable recommendations. For AI feature evaluation, every weak or violated case includes a clickable link or screenshot — no text-only claims.
Sara files bug reports directly to Jira via MCP — using the workspace's configured Jira project key, issue type, fix version, and priority. She attaches annotated screenshots, updates existing tickets with reproduction evidence, and never duplicates issues already in the backlog.
After meaningful runs, Sara persists experience under .sara/experience/runs/
and promotes stable lessons into heuristics or known issues. The knowledge base grows
with every session — making Sara progressively better at understanding the product,
its patterns, and its risk areas.
A complete reference of Sara's capabilities across modes, platforms, and evidence types.
| Capability | Mode | Platform | Output | Evidence Standard |
|---|---|---|---|---|
| Pre-build idea challenge | DISCOVER | Any | DISCOVER Review + Risk Matrix | Business docs, story, brief |
| Cycle brief / flow bet review | BET REVIEW | Any | Decision-grade QA artifact | Bet file, scope doc |
| User story readiness review | STORY REVIEW | Any | Story Review Report | Jira ticket, Quality Canvas |
| Bug reproduction — web | TEST | Web | Reproduction report + screenshots | Fresh live run required |
| Bug reproduction — mobile/Android | TEST | Mobile | Reproduction report + device screenshots | Fresh live Appium run required |
| Business-value journey test | JOURNEY | Web | Journey report + step evidence | Headed browser, real navigation |
| Journey + performance metrics | JOURNEY PERF | Web | Perf report + Core Web Vitals per step | Headed browser + DevTools timing |
| Page / feature bug sweep — web | SWEEP | Web | Sweep report + Jira tickets | Headed browser, annotated screenshots |
| Page / feature bug sweep — mobile | SWEEP | Mobile | Sweep report + Jira tickets | Appium, device screenshots |
| Structured Jira bug filing | REPORT | Any | Jira issue + annotated screenshots | Described or confirmed bug |
| Business risk report from Jira | RISK | Any | Risk report + ship recommendation | Jira story/epic + open bugs |
| Product Kit creation / update | PRODUCT KIT | Any | Quality Canvas + UJM + PM Dashboard | Business docs, Lean Canvas, codebase |
| Living Quality Canvas | QUALITY CANVAS | Any | Quality Canvas document | Lean Canvas answers first |
| GenAI feature evaluation | EVALUATE GENAI | Any | AI Quality Canvas + HTML report | DB cases (min 10), live sweeps (min 5), or files |
| Value-Quality Map for PMs | VALUE QUALITY MAP | Any | Value-Quality Map document | Quality Canvas + QA memory |
| Go / No Go release decision | GO / NO GO | Any | GO / NO GO verdict + conditions | Reports, Jira, or KB artifacts |
| RTL / bilingual defect detection | SWEEP / JOURNEY | Web | RTL defect list in sweep/journey report | Headed browser + locale switching |
| Persistent QA knowledge base | All modes | Any | Growing .sara/ KB | Auto-updated after every meaningful run |
Sara maintains a persistent, structured knowledge base in every project she works in — so each session starts smarter than the last.
Sara connects to the tools the team already uses — no new platforms, no manual hand-offs.
The beliefs that guide every run, every report, and every recommendation Sara makes.
A green test suite that ships a broken user journey is a failure. Sara measures quality by whether the product delivers what it promises to users — not by how many tests pass.
Every finding Sara reports is backed by real evidence — a screenshot, a link, a live reproduction step. Text-only claims without evidence are not filed as bugs. Period.
The cheapest defect to fix is the one you never build. Sara's DISCOVER and BET REVIEW modes exist to catch broken ideas, misaligned scope, and premature solutions before a single line of code is written.
Prior run history helps Sara plan — but it never serves as verdict evidence. Every TEST, SWEEP, and JOURNEY execution is a fresh live run on the current state of the product.
Sara's reports are written for product managers, not engineers. Risk is expressed in business impact language. Recommendations are actionable. Every output can be handed directly to a stakeholder.
Sara gets smarter with every run. Durable lessons become heuristics. Recurring bugs become known issues. The project's quality profile builds over time — so future Sara sessions start with real context, not a blank slate.