Automated Stage-Gate Verification Script

Problem/Feature Description

A machine learning platform runs a model training pipeline where each stage (data preparation, feature engineering, model training, evaluation) writes an output artifact to disk before the next stage begins. The platform currently has no automated stage-gate checks — engineers manually inspect files before approving the next stage. This approach doesn't scale as the team grows.

The engineering lead wants automated verification scripts that can be run as stage gates in CI. For each stage transition, the script should read the output artifact from the previous stage and confirm it is safe to proceed. The artifacts are JSON files. The platform runs on Python 3.8+.

Output Specification

Produce the following files:

1. stage_gate.py — A Python verification script that accepts two arguments: a file path and an expected SHA-256 checksum string. The script should:

   • Verify the file exists and is non-empty
   • Parse it as valid JSON
   • Verify the file's SHA-256 checksum matches the provided expected value
   • Verify any timestamp field in the JSON is within 5 minutes of current time (if the field exists)
   • Print a per-check result to stdout
   • Exit 0 only if all checks pass; exit 1 otherwise

2. contract.md — A one-page contract document describing the invariants this script enforces, with a table showing the boundary point, invariants checked, and failure action for each check.

The script should be runnable with: python stage_gate.py <artifact_path> <expected_sha256>