A script task's output

A script task runs a program you name, on its schedule or by hand, with no AI tool involved. HarmFlow shows what the program writes while it runs, and keeps all of it in the execution's log, so you can read it again once the execution ended.

While the program runs

Every line the program writes to its standard output or its standard error reaches the dashboard while the program runs. Each line is marked with the stream it came from, so you can tell the program's errors from the rest of its output. An execution you have open adds each line as it arrives: there is nothing to reload.

What the log keeps

The execution's log keeps every line in full. The lines of one stream stay in the order the program wrote them. Across the two streams, the lines stay in the order HarmFlow received them, and the dashboard and the log show that same order, so what you followed while the program ran is what you read in the log afterwards.

The start comes first

The execution's start is always shown before the program's first line of output, even for a program that writes the moment it starts. Anything it writes before its start is recorded is held back and shown right after the start: it is never dropped.

A busy program is never stalled

A program that writes a great deal at once is not slowed down or stopped by the recording of its output. HarmFlow releases what it holds back even when it cannot read the program's process identity or cannot save the execution's record, so the program always runs to its own end or to the task's timeout.

Secrets are redacted

A script's output is redacted the same way as an AI tool's output. A GitHub token in a line the program writes shows as [REDACTED], both in the dashboard and in the log: the token itself is not kept.