Runaway Guard
Enforces cost-safety discipline for paid AI and inference APIs by requiring explicit spending caps and execution bounds.
Runaway Guard is a specialized safety skill designed to prevent catastrophic 'bill shock' when working with paid AI services like Anthropic, OpenAI, and Fal.ai. It introduces cost as the third complexity dimension alongside time and space, ensuring that every API call site is protected by a documented 'wallet invariant.' By mandating pre-written cost contracts, concrete iteration bounds, and synchronized provider-side hard caps, this skill transforms risky unbounded loops and retry patterns into predictable, budget-safe implementations. It is essential for developers building autonomous agents, background jobs, or high-volume inference pipelines where a single logic bug could otherwise result in thousands of dollars in unintended charges.