← sources

Agents self-report "this is done and all tests pass" how do you verify that? How on agent scale?

source: reddit (reddit:AI_Agents)published 2026-10-12 13:46ingested 2026-10-13 02:36

retained text

We do the work at agent scale, and verifications also are done at agent scale. Do we verify if the agents do everything they claim? Or we just say "whatever" and merge it to main? 

For many of us, we definitely don't take agent's word for granted. For backend work, esp around client deliverables, I don't trust on status codes. It says 200 passed, but from a long chat, agent missed that some schema validations failed, and frontends has elevated errors. 

We all know agents shouldn't be grading their own work, we always have QA teams on top of devs, same way, something outside of agent should check the same thing and record what was called, what came back and what was compared. Something that can become **solid evidence over ephemeral test runs.** 

  
For people running agents in real workflows:

\- Do you have any external verification step, or do you rely on the agent's own test runs?

\- What would that evidence need to contain for you to trust it?

  
(Disclosure: I am building a tool, internally this tool becomes a fodder for my release notes bot for frontend and upstream subscribers. Happy to share in comments if anyone wants it)

document fetched 2026-10-13 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1x42atw/agents_selfreport_this_is_done_and_all_tests_pass