Reference

Glossary

Every term SQUAT uses, in one place. Cross-references point to the chapter with the full story.

Acknowledgment: The note folded into a returning persona's briefing when an issue they found has since been addressed — “the team fixed the save bug you found.” See Running rounds.
Actor: The pass that performs a persona in a task or discussion. Its SQUAT-generated bundle contains the actor briefing and current turn, never persona provenance. See How testers work.
Alter-ego: A persona's non-obvious second identity — the accountant who's a weekend musician — recorded only if the operator actually knows it. Explains behavior and delight the day job doesn't predict; surfaces occasionally in rounds, not constantly. See Recruiting.
Archetype: A persona's self-image — “The World-Builder”, “The Scene Purist” — the part of the identity triple that explains why they do things, distinct from the practice (the how). See Recruiting.
Backfill: Importing a historical round into the ledger at its original date. Backfilled rounds are marked, and count fully in every trend. See Importing & content packs.
Blocked: A result status: the attempt couldn't proceed at all (environment down, no login session prepared). Never counted as an app failure by itself. See Findings & scores.
clientInfo: The tester's stack recorded on every round — platform, model, plugin version. One half of “did the app change or did the tester change?” See Reviewing history.
Context (project context): Everything the squad should know about the product itself — what it is, who it's for, how to reach it, known limitations. Versioned; folded into every briefing. See Core concepts.
Content pack: A versioned set of optional personas, scenarios, discussion guides, frameworks, or squads that can be previewed and installed into a workspace. The free tutorial is the first pack; paid pack commerce is separate. See Importing & content packs.
Core features: Features you declare as the product's core loop on a project. Findings that land on them get double weight in the priority list. See Findings & scores.
Coverage: The per-feature record of what an attempt actually touched: not exercised, encountered, attempted, unexpected behavior, success. See Findings & scores.
Credential rule: No persona reads, types, prints, or pastes a password — ever, in any client, at any tier. Sessions are prepared before the actor starts. See Running rounds.
Default squad: The saved line-up a project authorizes for its rounds, panels, and steward work. A run may use all members or an explicit subset, never someone outside it. See Running rounds.
Deletion certificate: The content-free completion record for an authorized deletion request: scope, dates, counts, and outcome, without a copy of the deleted content. See Maintenance.
Dialogue ledger: The canonical, ordered event chain for a new round's spoken questions, probes, answers, hesitations, interjections, and disclosures. SQUAT reconstructs one stable, aligned dialogue from the inline turn ledger.
Depth check: The grading pass that tells rich, interview-grounded personas apart from thin, document-generated ones — run after imports, on request, and during maintenance sweeps. Classifies each profile recruit-grade, serviceable, or thin, and offers a short enrichment interview as the fix. See Recruiting.
Design panel: A qualitative, hypothesis-generating round in which personas answer one required, versioned moderator instrument about a concept or artifact. Never enters the scored trend. See Steward & panels.
Disposition: Your recorded call on a tracked issue: fixed, planned, investigating, wontfix, thanks, or deferred. Always your decision, never the AI's. See Findings & scores.
Dossier: The one-card summary of a squad member: who they are, their goals, friction behavior, and leave/delight triggers. See Recruiting testers.
Evaluator: The separate pass that judges a transcript without persona provenance until scoring is complete. See Running rounds.
First-person human: A persona source basis meaning the represented human supplied the material. Operator recollection about someone else is documentary, not first-person human.
Failure class: What actually failed on a failed attempt: app (your product), test (the scenario or persona definition), or agent (the tester/harness itself). The most valuable signal in the ledger. See Findings & scores.
Finding: One observed problem in one attempt: feature, title, severity S0–S3. Findings auto-become tracked issues. See Findings & scores.
Inconsistency: A result that may be transient or unreliable and should be re-run before it is treated as a regression. See Findings & scores.
Harness notes: The workspace's operating lessons about running tests themselves — an inconsistent scenario result or a slow staging build. Distinct from findings (about your product) and learnings (about a persona). See Running rounds.
Integrity check: A report-only scan for loose ends — issues whose evidence sits in the trash, references to rounds or personas that no longer resolve. It never changes anything; it just tells you. See Maintenance.
Intent registry: Your durable rulings about what each surface or behavior is for — the steward's ground truth. Rulings are superseded, never edited. See Steward & panels.
Invalidation: Flagging a compromised result or round with a reason so it's excluded from trends while remaining visible. There is no un-invalidate — and it's not a delete; for actually removing a round, see Trash. See Findings & scores.
Issue: The persistent, cross-round record of a problem, opened automatically by findings. Lifecycle: open → fixed (claim) → verified-fixed, or regressed. See Findings & scores.
Learning: A short behavioral observation a persona accumulates from rounds, making it act like a returning user. You pin, prune, or accept them. See Recruiting testers.
Ledger: Your longitudinal record of every round, result, and finding, in true chronological order. The point of the product. See Reviewing history.
License key: The access credential (starting sqk_) that connects your AI client to your workspace. It is not your workspace content-encryption key. A lost raw value cannot be reconstructed, but it can be replaced without losing workspace data. Always required; free tier available; three devices per key. See Setup & connection.
Live panel viewer: A first-person rich view of moderator questions, follow-ups or redirects, and panelist replies as they enter the canonical dialogue. It is a view of the ledger, not a separate transcript. See Steward & panels.
Mission: Your specific question for one round — distinct from the reusable scenario. SQUAT protects it automatically and shares it only through authorized round work. See Running rounds.
Model policy: An advisory per-project preference for which model plays actors, evaluates, or interviews. Honored where the client can; the round always records what actually ran. See Running rounds.
Moderator instrument: A versioned scenario for concept interviews/design panels. Spoken questions are stored separately from moderator-only objectives, probes, and listen-fors.
Narrative round: A round run with no browser tool connected: the tester reasons through the scenario against the product as described and says so plainly — useful for flow and copy critique, but it can't find what it can't see. The ledger records which mode ran. See How testers work.
Persona: See Squad member.
Playbook: A staged research program SQUAT builds for you from what you already have — pasted evidence such as support emails, NPS exports, error logs, or prior research, a short interview about what you want to learn, or both. It contains research questions grounded in that evidence, a recommended tester cast, the scenarios and instruments to create, a staged sequence of rounds, the decisions only you can make, and a budget posture. The playbook proposes; you decide what runs. Ask for one in plain language, or use /playbook with the plugin.
Provenance-excluding role bundle: A SQUAT-generated actor, moderator, or evaluator bundle that omits persona provenance and fields belonging to the other roles. SQUAT records which bundle it returned; it does not attest that an external model host isolated contexts, so the coordinator still uses separate model calls or contexts.
Practice: A persona's competence and context — “API/Developer”, “commercial photography” — the part of the identity triple (name, archetype, practice) that explains how they work. See Recruiting.
Priorities: The computed worst-first list of a round's findings: severity × distinct personas affected × core-loop weight, top ten. See Findings & scores.
Project: The record of one product under test. Carries context, a default squad, a default mission, model policy, and core features. Login secrets stay local; scenarios carry non-secret authentication policy. See Core concepts.
Research report: A versionable narrative synthesis for valid panel records or research-oriented rounds. Every saved revision or addendum is a complete append-only version; earlier versions remain retrievable.
Reference code: An optional readable two-word alias for a study, panel, or scenario. It helps conversation and lookup; the opaque ID and exact version remain authoritative.
Purge: The resumable irreversible hard delete of exact rounds already in the trash. SQUAT inventories the targets, then displays a short-lived 6-digit confirmation code only on the workspace owner's signed-in account page — never in a tool response — so a human must read it in the browser and type it back. On confirmation SQUAT locks the unchanged targets, removes referenced round data and artifacts, and checkpoints every step. A partial purge resumes under the same id. See Maintenance.
Regressed: An issue status meaning a claimed fix didn't hold — the problem came back. The loudest alarm in SQUAT; always sorted first. See Findings & scores.
UAT Report: The third analysis artifact. A Round Summary covers one round; a Research Report answers a question; a UAT Report assigns work — a run's findings, worst first, cited by issue rather than restated, with the true total always stated beside what is shown. See Reviewing history.
Report card: Unrelated to the UAT Report above — this is the steward's graded summary of a sweep: one A–F grade per heuristic, an overall grade set by the worst-informed area (never averaged), and trend arrows against the previous sweep. See Steward & panels.
Round: One testing session: squad members × scenarios, acted, evaluated, and recorded. See Running rounds.
Scenario: The tester's reusable task, stated as a goal in the user's own terms — never UI steps. The stable instrument that makes trends comparable. A scenario can live in one project (its home — typical for task scenarios written against a specific product) or in the workspace's shared library (typical for reusable discussion guides); a project sees its own scenarios plus the shared library. Running another project's scenario in a round is allowed — SQUAT just notes it, since comparing two products against the same goal is a legitimate method. See Core concepts.
Scores: Five per-attempt dimensions — discoverability, comprehension, task success, effort, confidence/trust — each rated 1/3/5, never blended. See Findings & scores.
Seat: Your membership slot in a workspace: role and your own message-acknowledgment state. Every license key ships with at least one seat. Hosted content protection is automatic and separate from seat administration. See Roles & security.
Severity (S0–S3): A finding's impact, from S0 (showstopper) to S3 (minor), set from observed impact — never from persona rhetoric. See Findings & scores.
Squad: A named, reusable line-up of squad members — “the mobile squad.” See Recruiting testers.
Squad member: A tester persona that may represent a human, be deliberately synthetic, or combine permitted sources. Also called a persona. See Recruiting testers.
Source basis: What contributed to a persona: first-person human, documentary, synthetic, or unclassified. Combined is a derived label. Separate from acquisition origin (interview/template/import/unclassified).
Synthetic audition: Synthetic sample responses used to calibrate a communication range. Never evidence, and never promoted into a persona source.
Starter: The free tier: three hosted squad members, one of which keeps learnings between rounds, one project, and 90 days of ledger history — with the full methodology at identical quality to every paid tier. The paid tiers are Solo, Team and Studio; they differ in capacity, never in method. See Setup & connection.
Learning slot: A squad member whose observations carry forward between rounds. On Starter, one of your three members holds it and you can move it whenever you like; on every paid tier all members keep their learnings.
Steward: The expert review that acts as your proxy — judging the live product against your intent registry and prescribing fixes. Not a persona. See Steward & panels.
System message: A notice SQUAT delivers at the start of a session — a broadcast announcement (maintenance, for everyone) or an account message (welcome, billing, for your workspace). Two choices, always: remind me later, or dismiss. Your client tells you which scope the choice affects. See Setup & connection.
targetBuild: Your product's version/build/commit, optionally recorded on a round — the other half of “did the app change or did the tester change?” Concept panels normally omit it. See Reviewing history.
Thinking style: An optional, operator-supplied read on how a persona approaches problems — analysis-first, steps-first, people-first, or play-first — brought from any framework (HBDI, Big Five) or plain gut feel. Never appears in the actor's script as labels or scores; it's translated into concrete behavior. Also powers the roster coverage check. See Recruiting and Maintenance.
Trash: Where deleted rounds go instead of vanishing: normally recoverable for 30 days, invisible in trends and portable exports, and included in an explicit full backup. A round becomes restore-locked when its prepared purge is authorized. Deleting is a lifecycle call; compare Invalidation, the data-quality one. See Maintenance.
Tutorial pack: The free optional practice content pack with a governed persona, squad, task scenario, and discussion guide. It is previewed and explicitly installed; it is not silently added to every workspace. See Quick start.
Validation interview: A short pass confirming an imported persona against your own read of that user, before its first round. See Recruiting testers.
Verify round: A round that re-tests specific known issues — ideally with the same persona that found them — resolving each to verified-fixed or regressed. See Findings & scores.
Version history: Every save of a persona or scenario creates a new version and archives the prior one — nothing is ever overwritten. Any version can be retrieved in full, and the whole timeline listed. Rounds pin the exact versions that ran. See Recruiting testers.
Workspace: The container for your squad, projects, rounds, issues, and keys — what your license key connects you to. See Core concepts.