Write audit_brief(brief, resume) — the claim-by-claim audit that
runs on every AI-drafted screening brief before a human reads it.
For each claim dict (claim, receipt, label), print one line:
- receipt exists in
resume→SOURCED {claim} - receipt is
Noneand label is"inference"→INFERENCE {claim} - receipt is
Noneand label is"sourced"→VIOLATION {claim} — sourced with no receipt - receipt is not
Nonebut missing fromresume→VIOLATION {claim} — receipt {receipt} not on the resume
Use f"{tag:11} ..." padding (tag column is 11 wide, then a space).
After the loop, print violations: {n} and return n.
Expected output:
SOURCED Five years payroll tenure
SOURCED Multi-state payroll at scale
VIOLATION Strong leadership presence — sourced with no receipt
INFERENCE Likely open to hybrid schedules
VIOLATION Workday certified — receipt L9 not on the resume
violations: 2
(Match it byte for byte — the dash before "sourced" / "receipt" is an
em-dash, —, not the hyphen key. Copy it from the block above.)