0%

Verification Coverage

Goal

Script-owned dual-exec parity + leak + ARC/codegen verification per program, results recorded through the §01 API.

Implementation Sketch

Scripts own every mechanical action; the LLM authors programs/tests and analyzes only. All mechanical work is a single python -m scripts.rosetta_test_runtime <subcommand> invocation directed by a work_item body line.

Work Items

  • Implement verify.py — dual-exec parity (interpreter output == AOT output), ORI_CHECK_LEAKS=1 leak check, ARC/codegen verify; each returns a structured pass/fail.
  • Record verification results into manifest.verification via the §01 API; pytest covers each check’s pass/fail mapping.