Test, upgraded: run it, then say "test this" (plugin 8.1.0)

The Hands-on AI plugin is at 8.1.0, and it changes how Step 5 works. A test round is now three moves: open the round in one chat, run each input in a new chat and say test this to grade it right there, and get the verdict in the last chat. No more copying the output back into the grading chat, and the skill checks the whole run, not just the final answer. The Test page has the new “How a Round Works” section, and the worked example shows a full round.

Deconstruct proposes the test inputs. In the last chapter of Deconstruct, you give one or two real inputs you’ve actually handled, and the skill proposes the rest, each aimed at something that could go wrong: the typical case, a hard case, an empty input, a must-never rule or a human gate, and an embedded instruction where your workflow reads external content. You correct the list. Golden examples are asked for only where a real past output exists.

Every workflow ends with a “What I did” list. Whatever Build creates now closes each run with a short list: the steps it took, where it paused for you and what you decided, what it did with your tools, and where the deliverable is. That is how Test can prove a gate paused or a rule held, on any platform.

Test grades where the run happened. The skill works out by itself whether the new chat can reach your results file; if it can’t, it tells you in one sentence to paste the run back into the chat where you opened the round. A guard refuses to grade any run that saw your design before it started. The report card lists output checks first, then rules, gates, and step outputs with evidence quoted from the What I did list. If you close your laptop mid-round, saying “test my workflow” in a fresh chat picks up at the next input.

Build and Improve respect an unfinished round. Build asks you to finish the round before it rebuilds anything, and Improve’s regression check uses the same run-then-test this pattern. Two new glossary entries cover Round (Test) and What I did summary.

Update from the plugin marketplace, or download the skill ZIPs.