Platform
What sets SkillCort apart
Anyone can show a quiz. SkillCort is built for the parts that are hard: authoring real work, protecting the session, keeping AI accountable, and turning scores into decisions you can defend.
Candidate experience & delivery
Decide who gets in — and how many chances they get
How a candidate reaches an assessment, how they identify themselves, and what happens when they come back — set per delivery, with room for one-off exceptions.
See Candidate Experience & DeliveryFive ways in
A public link, a unique per-candidate invitation link, a short access code, guest-list matching on one or two identifiers, or sign-in for your own organization members.
Ask nothing, or exactly what you need
A link with no access form starts the assessment immediately; add fields and candidates register with the details you asked for, kept on their record.
Attempt rules per delivery
How many attempts count, whether an interrupted session resumes, how long an attempt may stay open, and what a candidate meets when they return after finishing.
Retakes & extra time, one candidate at a time
Grant a retake to the person whose power cut ended their session, or extended time up to 3× as an accommodation — the cohort's rule stays exactly as it was.
Windows, expiry and caps
Open and close dates, an expiring public link, and a registration limit; candidates who arrive early or late meet a designed screen, not an error.
Access log before the exam starts
Registrations, code entries, failed matches and denials are logged with their reason and outcome — filterable in the app, exportable as CSV.
AI interviews & simulations
Run a conversation, then review the evidence
Use disclosed AI-led interviews and case simulations when the skill is best shown through dialogue. The conversation remains available for human review and scoring.
Explore AI Interviews & SimulationsAI-led interviews
Candidates answer spoken interview questions while the interviewer responds turn by turn in a structured session.
Role-play simulations
Put candidates into a guided case conversation so judgment, communication, and follow-through become observable.
Avatar or voice-only delivery
Choose a talking interviewer avatar or a voice-only presentation for each live interview or simulation.
Reviewable conversation record
Spoken responses are transcribed and kept with the session so evaluators can return to what the candidate actually said.
Disclosed before the session
Candidates see how AI and recording are used before they begin, alongside the capture settings for that assessment.
People score the outcome
AI conducts the configured conversation; evaluators apply the rubric and remain responsible for scores and decisions.
Anti-cheating
Integrity in layers, not a single lockdown switch
Enable proportional browser, capture, answer, and cohort layers per delivery. Signals land in a reviewable timeline, and configured capture is disclosed up front.
See Integrity & ProctoringIn-browser lockdown
Fullscreen, clipboard, print, second-display, and keyboard safeguards where supported by the browser — nothing for candidates to download. Sessions running in lockdown also carry an on-screen watermark with the candidate's identity and a live timestamp.
Honeytokens
A per-attempt marker can detect when protected question text returns inside an answer after leaving the exam — an exact round-trip signal, not proof of which tool was used.
Tamper-evident session log
Events are hash-chained in sequence; server-side re-verification reports the first detected break among the rows present.
Camera, screen & microphone
Photo or video capture with AI-assisted review that points a human reviewer at moments worth checking — AI never issues a verdict.
Behavioral signals
Large-paste detection, synthetic-typing dynamics, focus-loss correlation, VM detection, screen-share loss recovery.
Cohort analytics
Collusion signals across candidates — shared wrong answers and text similarity — plus a live monitor wall during sessions.
Governed AI
AI that assists — and can prove it never decides
AI can draft, transcribe, conduct configured conversations, and suggest scores; it cannot independently write a candidate score or decision.
Read the AI ManifestoHuman-in-the-loop by architecture
AI drafts, transcribes, flags, and suggests scores; a person reviews and decides. There is no automated candidate outcome.
Traceable AI surfaces
Registered generation surfaces record model and pinned prompt identity; live conversations keep a separate session-configuration ledger and transcript.
A public AI manifesto
What we build, what we refuse to build (no cross-client candidate memory, no hidden reputation scores), published for anyone to read.
Incident controls
Scoped incident controls block selected AI provider calls. Saved responses and non-AI work remain available; affected AI flows fail safely.
Candidate transparency
Before starting, candidates see configured capture controls and whether an AI-guided section is used; the public Trust Center publishes versioned governance statements.
Evaluation & evidence
An evaluator workflow, not a spreadsheet export
Scoring is a governed pipeline: a work queue, rubrics with AI suggestions, blind scoring, and a finalize lock that separates provisional from official.
See Evaluation & EvidenceEvaluator work queue
Every submission with its state — ready, in progress, ready to finalize, finalized — searchable, sortable, and built for teams.
Blind scoring
When enabled, written-work review replaces the candidate profile with a server-generated alias. Results and identity-bearing media remain outside that guarantee.
Rubrics + accountable AI suggestions
AI suggestions stay separate; evaluators apply a suggestion or enter their own criterion score. Changing a saved score requires a reason.
Finalization lock
Scores stay provisional until a human declares them final; the database itself refuses score writes on a finalized attempt.
Decision views with psychometrics
Decision and analytics views combine skill rollups and role fit with assessment-level alpha and ICC; blueprint evidence adds omega and group-disparity checks when enough data exists.
Branding
Your brand in front of candidates, not ours
Organization branding carries from invitations into reports, with per-assessment overrides for candidate screens and emails.
Explore the productOrganization branding studio
Organization logo, name, and accessible accent color across the candidate access and exam flow.
Per-assessment overrides
Campaign-specific logo, hero banner, page background, copy, and footer contacts — agencies can brand per client.
Custom email templates
Invitation and reminder emails with your subject, body, button, sender name, and merge variables — live preview included.
White-label reports
Candidate reports and answer-sheet PDFs open with your cover, not a vendor's.
Reporting
Evidence you can hand to legal, not just charts
Every number traces back to its evidence — and the exports are built for the day someone challenges a decision.
See Reports & AuditCohort & candidate reports
Item analytics (difficulty, distractors), reliability cards, per-candidate deep reports, and per-candidate answer sheets as PDF.
Audit pack
One JSON export with integrity status, event timeline, and a server-side hash-chain verification result for the rows present.
Evidence wallet
A per-candidate evidence file: outputs, scores, fit, integrity signals, and the decision — viewable and exportable as JSON.
Validation evidence
Blueprint-level reliability, observed criterion validity, and group-disparity checks when enough outcome and group data exists.
Open exports & API
CSV exports for candidates, scores, decisions, and skills; an API-key-protected results API for your own systems.
See it on your own use case
Bring a role or an exam you run today — we'll build it live in a demo.