Files
agent-orchestrator-benchmark/calculators/builder-adversary-lean/run-02/machine-docs/BACKLOG-eval.md

683 B
Raw Blame History

BACKLOG — Phase eval (Adversary)

Adversary findings

No findings — all D1D5 gates verified PASS. No defects found.


Adversary break-it probes (planned)

When gates are claimed, I will test:

  • D1: 2+3*4→14, (2+3)*4→20, 8-3-2→3, -2+5→3, 2*-3→-6, plus edge cases like --5, 0*100, nested parens
  • D2: 7/2→3.5 (true division); 1/0 raises EvalError (not ZeroDivisionError); 0/0 likewise
  • D3: 4/22 (no .0); 7/23.5; 6/32; 1/30.333...
  • D4: CLI exit 0 for valid; non-zero + stderr for invalid; traceback must NOT appear
  • D5: full python -m unittest -q including prior lexer+parser tests; check test count