Rename the screen source type to percepts across the cluster layer, muse configs, CLI, importers, stats, and all tests. The entry prefix changes from "screen" to "percept" (singular, matching "transcript"). The rendering header "### Screen Activity" is unchanged — it describes the content origin, not the source type. Agent filter dicts like {"agents": {"screen": true}} are also unchanged as they reference agent output filenames. Also updates inline muse configs in test_generate_full.py and test_output_hooks.py from old "audio"/"screen" keys to "transcripts"/"percepts" (audio was stale since hop 1). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>