What evidence proves an agent finished the task? Reddit debates completion checks
Normal_Succotash_520 · reddit · 2026-09-16
A Reddit discussion tackles how to define task completion for agents that mutate external apps: a successful write call doesn't guarantee the CRM record is correct, or that a later workflow won't overwrite it.
The author leans toward a short checklist tied to the user's request — record exists, expected field values, a separate receipt for notifications — with each item requiring evidence fetched from the destination app before the agent reports done. The tradeoffs are extra reads, added latency, and deciding how long to wait on asynchronous changes.
Questions posed to the community: do you define checks per workflow, rely on tool response contracts, or add a separate verification step — and what do you deliberately leave unchecked?
More from coding & agent
- Sistava launches AI employee platform for business workflows, plans from $25/mo — Mahmoud_Zalt · 2026-09-16
- Debate: Is context rot a hard ceiling for LLM agents, or a solvable problem? — binarybits · 2026-09-16
- Does anyone actually use Codex ultra mode? Subagents just produce 'a mountain of slop' — wstone_bd · 2026-09-16
- Same function runs 14x slower in production: 500ms locally vs 7000ms in cloud — DanielLockyer · 2026-09-16
- Context rot vs. agent optimism: a debate over whether LLM agents can ever run unsupervised for days — michaelbd · 2026-09-16
- Six citation drifts surfaced after 5 days — record source URL and quote or don't cite — Agent-OmegaLT · 2026-09-16