verification-oraclelisted

# Verification Oracle Deterministic re-execution of a command/code task gate: run the declared check again, compare the actual result to what the gate expects, and return an objective pass/fail with the evidence that produced it. This is the mode a verifier loads when a gate's `verification_type` is one of the two DETERMINISTIC types in `gaia.state.VALID_VERIFICATION_TYPES` -- `command` and `code` -- as opposed to `semantic` (needs human/rubric judgment) or `self_review` (trusts the producing agent's own statement). ## Position in the flow This mode runs mid-verification, not standalone. Upstream, a planner already authored the gate (`gaia task gate add ... --type=command|code --evidence-shape='<check>'`, persisted in `task_gates`) or a producing agent already proposed `evidence_report.verification` on its contract (`gaia.contract.validator`) -- either way the check spec exists as text before this skill runs; the oracle never invents what to check, only whether it currently holds. Downstream, the pass/fail verdict feeds whatever decides the gate's/AC's status next -- so a wrong verdict here propagates as a false pass or a false block one layer up, not just a local mistake. ## Why command and code are ONE mechanism, not two Read `gaia.state.gate_validation` and `gaia.contract.validator` before assuming `code` needs different execution machinery than `command`: both types resolve to the SAME shape, a runnable string -- `evidence_shape` on the persisted gate, or `command` o