T15: no baseline exists — the run should build one itself

[aleqsio]

Points at screenmaps_branch: screenmaps-t15, which does not exist, so this PR has nothing to diff against. The workflow carries actions: write.

Expected: the run detects the missing baseline, starts the baseline workflow on main itself, and comments "building the first map" rather than telling a human to go and click it.

Not covered: the closing half of the loop. The baseline workflow on main publishes to the real screenmaps branch, not to screenmaps-t15, so the automatic re-run cannot find a map afterwards. Testing that end to end needs the real branch deleted.

🤖 Generated with Claude Code

[github-actions]

🗺️ screenmap · building the first map

This repo had no baseline to compare against, so one is being built from the default branch now. It maps every screen, so it takes longer than a normal run — roughly 20 minutes.

This PR is mapped automatically as soon as it finishes. Nothing for you to do.

_{Follow the baseline run →}

[aleqsio]

Closing to stop an auto-baseline dispatch loop: this PR points at screenmaps-t15, which the baseline workflow never publishes to, so each completed baseline would re-run this PR and trigger another. The dispatch behaviour under test already passed. Loop guard being added separately.