Know who actually did the work
Remote pipelines face proxy candidates and identity fraud. SkillFoundry turns behavioral signals into integrity review prompts — decision-support with a human in the loop, never an auto-reject.
How it works
Provenance built on behavioral signal
The same event stream that powers scoring becomes tamper-evident evidence — without webcams, biometrics, or keystroke content.
Tamper-evident trail
Every session is hash-chained and signed. Post-hoc edits are detectable — the evidence trail holds up when a hiring decision is questioned.
Behavioral authorship signature
A behavioral signature of how the candidate works — activity, not content. No webcams, no keystroke logging.
Human-in-the-loop review
Signals surface to reviewers as decision-support. Nothing auto-rejects. Every confirm or dismiss requires a documented reason.
Prompts you can understand
What gets flagged for review
Each prompt links back to the session evidence. Reviewers confirm or dismiss with a required reason.
Working pattern changes
Part of a session looks different from the rest of the candidate’s work. Reviewers check the replay before drawing conclusions.
Work from outside the session
Some submitted code may not have been written in the session. Reviewers confirm with the replay and test history.
Differences across stages
A later stage looks unlike the candidate’s earlier sessions. It prompts a closer look, never a verdict.
For your organization
How authenticity helps hiring teams
Integrity signals complement orchestration scores and benchmarks — they answer 'was this work actually done by this person?'
Defeat proxy candidates
Differences across sessions prompt review when the person doing the work may not be the person you are hiring.
Review prompts, not verdicts
A flag is a prompt to investigate — paired with session replay and test results, not an automatic rejection.
Protect candidate trust
Candidates are never shown raw flags as accusations. Reviewers document reasoning; dismissed signals stay on the audit trail.
Demo mode
This page uses canned integrity data in a simplified version of the panel reviewers see on submission detail.
What you are looking at
Panel 1 shows a session with two items waiting for review. Panel 2 shows an item a reviewer dismissed, with their notes. Panel 3 shows a clean session with nothing flagged.
Panel 1
Signals needing review
Alex Chen's pagination session — two items flagged for human review.
Integrity review
Review prompts only — not an accusation and never an auto-reject. Reviewers confirm or dismiss each one with a documented reason.
Working pattern changed mid-session
Needs reviewPart of this session looks different from the rest of the candidate’s work. Check the replay before drawing conclusions.
Code may not be the candidate’s own
Needs reviewSome of the submitted code may have come from outside the session. Confirm with the replay and test history.
Panel 2
Reviewed and dismissed
A difference from earlier stages, reviewed with documented reasoning — the audit trail preserves the decision.
Integrity review
Review prompts only — not an accusation and never an auto-reject. Reviewers confirm or dismiss each one with a documented reason.
Differs from earlier stages
DismissedThis session differs from the candidate’s earlier sessions. Use it as a review prompt, not a verdict.
Reviewer note: Replay shows a consistent problem-solving approach; the difference is explained by a long debugging pause.
Panel 3
Clean session
Nothing flagged for review.
Integrity review
Review prompts only — not an accusation and never an auto-reject. Reviewers confirm or dismiss each one with a documented reason.
Nothing flagged for human review on this session.
Ready to add provenance to your hiring pipeline?
Integrity signals run automatically on every instrumented session — reviewers decide what to act on.