validate_reputation_scoring: documented 'cheaters lose score' property is never enforced (dead loop)

[SwasthikaDev]

Summary

validate_reputation_scoring documents the property "reputation scores decrease for agents that cheat", but that property is never asserted. The scores dict is accumulated and then read only inside a loop whose body is pass, so the score arithmetic is dead code. The validator's one real check is a set comparison between cheat: senders and report:...:bad recipients, which passes vacuously when the trace has no cheat: lines.

Low severity (the validator still does something, and the reputation scenario's own discrimination works), but the stated invariant isn't enforced, so a trace that violates "cheaters lose score" is not caught.

Where

packages/nest-core/nest_core/validators.py, validate_reputation_scoring (~lines 787-847). The dead loop:

for cheater in cheaters:
    if scores[cheater] >= 0:
        # ... comment ...
        pass

scores is never compared against anything after this.

Reproduction

from nest_core.validators import validate_reputation_scoring

def send(a, to, msg): return {"ts": 1.0, "agent": a, "kind": "send", "to": to, "msg": msg}

# agent-9 receives 5 "bad" reports (its score should fall) and there are no cheat: lines
events = [send("reporter", "obs", "report:round:agent-9:bad") for _ in range(5)]
print(validate_reputation_scoring(events)[0].passed)   # True

The validator passes regardless of the accumulated scores.

Suggestion

Either enforce the documented property (assert that agents with net-negative reports end below their starting score, or below non-cheaters), or drop the scores accumulation and the dead loop and tighten the docstring to match what the check actually verifies.

Reproduced on main (latest as of filing).

[IlIGODIlI]

Hi! I'm interested in working on #98. Before I start implementing a fix, I wanted to clarify the intended invariant. The current validator computes reputation scores but only has access to the final score. Should the validator enforce that every bad report decreases the running score, or is the intended property that cheaters eventually end with a lower final score (or below a threshold)? I want to make sure my PR matches the expected behavior.