dont allow all 0

[xzrderek]

Note

Enforces a small success threshold for evaluation tests (host/Docker), forwards EP_SUMMARY_JSON into containers, and adds tests validating threshold behavior.

• CLI:
  • Local test runner (eval_protocol/cli_commands/local_test.py):
    • Always pass --ep-success-threshold 0.001 to pytest (host and Docker) so all-zero-score runs fail.
    • Forward EP_SUMMARY_JSON into Docker when it points under ~/.eval_protocol to expose summary artifacts on host.
  • RFT creation (eval_protocol/cli_commands/create_rft.py):
    • Expand validator docstring to clarify enforced success threshold behavior.
• Tests (tests/test_evaluation_postprocess.py):
  • Add threshold tests asserting all-zero scores fail at success=0.01 and equality passes.
  • Import updates to support new tests.

^{Written by Cursor Bugbot for commit a891471. This will update automatically on new commits. Configure here.}