Agent approvals work better when the checklist lives outside the prompt
bolerbox · reddit · 2026-07-22
The author argues that agent approval rules work better when they live outside the prompt. Prompt text can drift, be edited, inherit context, or be bypassed by a rushed operator, so the approval gate should be a separate, human-readable checklist.
The suggested checklist includes:
- what source material was used
- what the agent is allowed to change
- what needs human approval
- what counts as a failed run
- where decisions and logs are saved
The post says this keeps the prompt flexible while making the approval layer stable and easier to debug. It also warns against making the checklist too smart: if it turns into another fuzzy evaluator, the agent is effectively grading itself again.
More from coding & agent
- Hugging Face wraps 6-month Training Agents series with open recordings and code — SergioPaniego · 2026-10-26
- Agent verified 'success' after double-voting: a lesson in harness vs model bugs — TejasKumar_ · 2026-10-26
- diginomica: agentic automation hype is overdone, and 'The Hackening' of enterprise security has begun — jonerp · 2026-10-26
- 'Vibe verification': ~60% of AI-written formal specs prove the wrong property — basedjensen · 2026-10-26
- Devin launches Security suite powered by Claude Mythos with $100M+ credits — DevinAI · 2026-10-26
- HackBench injects reward hacks into agent trajectories to stress-test monitors — DanielKhashabi · 2026-10-26