A token-bucket rate limiter is exactly the kind of well-scoped coding task AI handles well.

Token-bucket rate limiting is a standard algorithm with a fixed structure: refill tokens at a steady rate, consume one per request, block or queue when empty. The pattern is the same regardless of which third-party API is being throttled, making this highly repeatable.

Success criteria are concrete and testable: the limiter must not exceed 100 requests/minute, distribute requests evenly, and integrate with the existing queue consumer. These can be verified with unit tests and load simulation, leaving little room for subjective interpretation.

The algorithm itself needs no external data, but correct integration requires access to the existing queue consumer code. Without that context, the agent can produce a correct standalone implementation but may need a human to wire it in precisely.

A buggy rate limiter could still trigger API rate-limit errors or, worse, silently drop requests — both are recoverable but disruptive. The risk is moderate: errors surface quickly in testing and don't cause irreversible data loss, but a production bug could cause service degradation.

The design choices here — bucket capacity, refill interval, burst handling — are well-understood engineering trade-offs with documented best practices. No taste, ethics, or relationship context is needed; a competent agent can reason through the options and justify its choices.