feat: checkpoint device agent runtime milestones

research/001-scene-model/README.md

@@ -0,0 +1,5 @@
+# Scene Model
+
+This track holds papers, experiments, and benchmarks about representing device
+screens as structured scenes. It is for research notes and evaluation assets,
+not runtime code.

research/002-world-model/README.md

@@ -0,0 +1,5 @@
+# World Model
+
+This track holds papers, experiments, and benchmarks about persistent state,
+environment modeling, and task-relevant memory across observations. It is for
+research notes and evaluation assets, not runtime code.

research/003-skill/README.md

@@ -0,0 +1,5 @@
+# Skill
+
+This track holds papers, experiments, and benchmarks about reusable task skills,
+skill authoring, versioning, and retrieval. It is for research notes and
+evaluation assets, not runtime code.

research/004-agent/README.md

@@ -0,0 +1,5 @@
+# Agent
+
+This track holds papers, experiments, and benchmarks about agent roles,
+delegation, critique, recovery, and collaboration. It is for research notes and
+evaluation assets, not runtime code.

research/005-memory/README.md

@@ -0,0 +1,5 @@
+# Memory
+
+This track holds papers, experiments, and benchmarks about task memory,
+episodic traces, long-term recall, and retention policies. It is for research
+notes and evaluation assets, not runtime code.

research/006-planner/README.md

@@ -0,0 +1,5 @@
+# Planner
+
+This track holds papers, experiments, and benchmarks about planning strategies,
+prompting, decomposition, and plan repair. It is for research notes and
+evaluation assets, not runtime code.

research/007-ui-understanding/README.md

@@ -0,0 +1,6 @@
+# UI Understanding
+
+This track holds papers, experiments, and benchmarks about semantic UI
+understanding, visual grounding, accessibility trees, and multimodal screen
+interpretation. It is for research notes and evaluation assets, not runtime
+code.