How long should an approval for an agent action stay valid?
nav8_ai · reddit · 2026-11-11
A Reddit post examines a key agent-design question: the lifetime of a "yes." The author's agent asks before sending or spending; currently each approval covers exactly one action — the exact message to the exact recipient — and dies if content changes or minutes pass. The alternative, broader grants like "you may email this client this week," are less annoying but the approved thing is no longer the thing that runs. Failure modes to guard against: the agent rewording a draft post-approval, or an injected line changing the recipient while the old approval remains valid. Do you scope approvals per action, per session, or per task?
More from coding & agent
- Plurai launches free agent simulation testing inside Claude Code and Cursor — PrajwalTomar_ · 2026-11-11
- PlayCanvas engine repo hits 10,000th issue/PR — and it's a Claude-authored merge — willeastcott · 2026-11-11
- AI-built game nerfed 40% after Claude playtest bot beat it in 6 minutes — builderjaydub · 2026-11-11
- Anthropic launches Claude Tag for Slack, claims 65% of its product code is AI-written — nikola_mr64990 · 2026-11-11
- The key to expert-driven evals: maximize information gain per unit of expert attention — danshipper · 2026-11-11
- Open-source node turns any ComfyUI workflow into a one-click model, 19 models bundled — jefoce · 2026-11-11