LEGAL-REVIEW-CHECKLIST.md markdown
42 lines 2.1 KB
Raw
sha256:ca1d0e687bff8686b37126eb8e4fb6d38b40e189352a4920d66ae89c7274e340 Add T4 job cancellation retry and validation Human minor ⚠ breaking 41 days ago

Legal Review Checklist

Simple Summary

No private learner data or paid GPU work can enter Scooling Lab until legal, security, dependency, and data controls are reviewed and accepted.

Pre-Private-Data Checklist

  • Confirm the repository license and distribution model for Scooling Lab.
  • Confirm every runtime dependency is listed in DEPENDENCIES.md.
  • Confirm every runtime dependency license is allowlisted or separately approved.
  • Confirm no AGPL-covered Studio or CLI path is imported, copied, bundled, or exposed.
  • Confirm dataset consent, scope, retention, export, and deletion policies are approved.
  • Confirm the API rejects browser-supplied worker URLs, callback URLs, shell commands, file paths, and unapproved model ids.
  • Confirm logs exclude prompt bodies, private notes, secrets, tokens, local paths, and raw payloads.
  • Confirm artifact retention and deletion rules are written and tested.
  • Confirm incident response ownership before any private-data pilot.

Pre-Paid-GPU Checklist

  • Confirm payer-visible cost policy and spending caps.
  • Confirm quota and replay controls for duplicate jobs.
  • Confirm cancellation and cleanup behavior before work starts.
  • Confirm GPU credentials are environment-scoped and not shared with the browser.
  • Confirm network egress rules are allowlisted.
  • Confirm no private data reaches test fixtures, logs, telemetry, or artifacts.
  • Confirm generated model artifacts have provenance, retention, deletion, and export records.
  • Confirm legal approval covers the exact package versions, model licenses, data sources, and deployment lane.

Required Seven-Tier Evidence

  • Unit tests for schema, state, license, and error decisions.
  • Integration tests for API-to-worker and BOM generation.
  • End-to-end tests for fixture job completion and artifact listing.
  • Stress tests for deterministic duplicate creation and queue limits.
  • Data-integrity tests for stable ids, stable hashes, restart persistence, and Unsloth evidence.
  • Performance tests for bounded status/list endpoints and BOM runtime.
  • Security tests for path traversal, command injection, URL injection, secret scanning, and AGPL path blocking.
File History 1 commit
sha256:ca1d0e687bff8686b37126eb8e4fb6d38b40e189352a4920d66ae89c7274e340 Add T4 job cancellation retry and validation Human minor 41 days ago