Handoff Pack
A focused interview brief that travels with the candidate to the next round: the decision so far, the evidence, the limits, and the questions worth asking.
Handoff Pack
The Handoff Pack is a single page that travels with the candidate to the next interviewer. It exists so that the next person in the chain can pick up exactly where the previous reviewer left off, without re-reading the entire report.
What is in the pack
| Section | What it is |
|---|---|
| Candidate | Name and email. |
| Decision so far | The current verdict, the reason code, the decided-at time, and whether the decision was automated (i.e. by policy, not by a person). |
| Evidence | Best score, effective recommendation, evidence state and finality, cohort, integrity flags, and a skills summary. |
| Limits | A short list of what the assessment cannot speak to. |
| Probes | Ordered questions the next interviewer should ask, with the basis (the stored fact that produced each one). |
| Probes omitted | A count of probes the model considered but did not include, so the reviewer knows the pack is curated, not exhaustive. |
How probes are produced
Probes are derived from the candidate's evidence. Each is always a question — never a finding — and each carries a stated basis: the measured fact that produced it. A probe phrased as a finding would have the reader adjudicating an accusation. A probe phrased as a question, with its basis visible, asks them to investigate.
Probes are ordered by how much a human is needed to resolve them:
| Kind | Meaning |
|---|---|
INTEGRITY | Resolve a flagged moment (e.g. "what was happening on screen at 12:34?") |
UNMEASURED | Cover a topic the assessment did not test (e.g. "how do you approach system design?") |
DIVERGENCE | Reconcile a difference between signals (e.g. "high score, low confidence — what was your reasoning?") |
WEAKNESS | Probe a recurring weakness (e.g. "you struggled with the recursion problem — talk me through your approach") |
PROCESS | Probe a process signal (e.g. "long deletion ratio on the second problem — what was the iteration like?") |
The omitted count is reported on its own so the reviewer knows the pack is curated, not exhaustive.
When the decision is empty
If nobody has decided on this candidate yet, the Decision section reads "nobody has decided yet" — not a fabricated placeholder. The pack is still useful: it contains the evidence, the limits, and the probes, which is the right input for the next reviewer.
When the decision is automated
If the decision was made by the policy rather than by a person, the pack records
automated: true and the policyVersionId. This is the explicit answer to
"was this an automated employment decision".
Curated, not exhaustive
Probes are the model's best attempt at the questions worth asking. They are a starting point for the interview, not a script. Replace, add, and reorder them to match your process.
Exporting and sharing
You can hand the pack to the next interviewer in two ways:
- Export as PDF — produce a portable, printable copy of the pack for an interviewer who is not in the dashboard.
- Share link — create a link you can forward to a colleague. A share link can carry an expiry, so it does not remain open indefinitely, and it can be revoked when it is no longer needed.
| Action | What it is |
|---|---|
| Export PDF | Downloads the pack as a document. |
| Create share link | Creates a forwardable link, with an optional expiry. |
| Revoke share | Disables a previously created link immediately. |
Both are available from the Handoff Pack, alongside the on-screen pack itself.