Knowledge Freshness Drift Checker

Scenario

Use before an agent is allowed to consume a knowledge release whose source documents may have aged or changed.

Input and output

Input is JSON with source records and a deterministic as_of date. Output contains FRESH, STALE, or UNVERIFIED flags, reasons, and freshness_hash.

Execution and recovery

Run python scripts/freshness.py examples/input.json. Set max_age_days in the input. Invalid dates and negative thresholds return visible errors. Fix metadata and rerun; do not silently treat missing verification as fresh.

Privacy boundary and acceptance

Use only source ids, verification dates, status, and redacted references. The sample must produce both a fresh record and a stale or unverified record with a stable hash.