← sources

A coding agent's "tests passed" message is still just a message

source: reddit (reddit:AI_Agents)published 2026-10-28 03:04ingested 2026-10-29 02:36

retained text

Microsoft's Agensh repo has a shared notes board for coding agents. One detail worth sharing: the board stores reported verification, but doesn't run those checks itself.

That matters when one agent is deciding whether to trust another agent's progress.

The board's README describes a few useful mechanics:

• Notes have types: FACT, OBSERVED, FAIL, CLAIM and PATCH\_SUMMARY.

• Most notes are capped at 100 characters; patch summaries at 300. Fuller detail can be fetched separately.

• A task claim expires from live reads after an hour by default, so an abandoned claim doesn't stay active forever.

• A matching FAIL invalidates that author's latest candidate patch.

• A new patch summary replaces the same author's earlier summary for the same work label.

But none of that makes a reported result true. The service checks the submitted text, not whether the verification command actually succeeded.

The useful distinction is between coordinating work and verifying work. A shared board can tell agents what someone attempted, claimed or reported. Before treating a patch as ready, the workflow still needs a way to check the result itself.

document fetched 2026-10-29 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1xhd02i/a_coding_agents_tests_passed_message_is_still