Prompt rules don't stop double payments: agent testing shows code and ledgers must enforce spend
Tiny_Associate_8254 · reddit · 2026-10-26
A developer building EmberTail, a test harness for money-moving agents, argues that spending controls like "never pay twice" don't belong in system prompts. In tests, an autonomous agent with a prompt-level no-double-payment rule still made three duplicate payments, all with the same pattern: the payment settled but the reply timed out or errored, and the model—unable to distinguish "call failed" from "call succeeded but reply lost"—reasonably paid again. The fix was boring: move the rule into code, holding any payment with an unknown outcome until a human checks. Zero doubles after that. His rule of thumb: the prompt decides whether the agent wants to pay, code decides whether it's allowed, and the ledger decides whether it happened. Even his well-behaved example agent still fails 11 of 63 test cases, underscoring how hard this is.
More from coding & agent
- Claude + HeyGen: building a creator video wall from just a few prompts — petergyang · 2026-10-26
- Fudan and Tencent unveil DraMaze, a multi-agent pipeline for diverse microdrama scripts — Siyang He · 2026-10-26
- Alibaba's MultiAgent-OPD trains hierarchical multi-agent systems end-to-end with on-policy distillation — TaoLiveAIGC · 2026-10-26
- Why Apple's folding iPhone shows black bars: apps built with older tools are deliberately boxed — aakashgupta · 2026-10-26
- License gotchas when picking an open source OCR model for commercial use — TangeloOk9486 · 2026-10-26
- Mistral ships Voxtral 3: speech models and Voice Agent SDK across 50+ languages — AxSaucedo · 2026-10-26