Bot comments nobody reads

It's tempting to automate process enforcement by having bots leave comments on PRs and issues. Missing a checklist item? Bot comment. Didn't fill out the template? Bot comment. PR sitting idle? Bot comment. The automation is easy to build and feels productive. But within a few weeks, the team learns to scroll past them.

The problem is that comments are advisory. They show up in the same stream as actual human feedback, and once someone learns they can merge anyway, the comment is just noise. Worse, noisy bots train people to ignore all bot output, including the comments that actually matter.

The fix depends on how much you care about the thing being checked. If it truly matters, make it a required status check that blocks the merge. If it can't be a status check, move it earlier in the process. GitHub's YAML-based issue forms can make fields required at submission time, so there's no issue to comment on in the first place. PR templates put the checklist in the author's hands instead of posting it after they've already written the description.

If something isn't important enough to block on and isn't important enough to build into the form, it probably isn't important enough for a bot comment either. Kill it and reduce the noise.

The one exception is scheduled housekeeping: stale issue warnings, idle PR nudges. These work because they're infrequent and they carry a consequence (the issue gets closed). But even these should fire sparingly. A bot that nags every three days is a bot that gets muted.