fix: give excalidraw skill its own section in agent system prompt

The excalidraw skill targets MCP tools (read_me, create_view) not the
widgetRenderer component, so it needs its own top-level section in the
system prompt rather than being nested under widgetRenderer instructions.

Author: jerelvelarde

apps/agent/main.py

@@ -10,10 +10,11 @@
 from src.query import query_data
 from src.todos import AgentState, todo_tools
 from src.form import generate_form
-from skills import load_all_skills
+from skills import load_all_skills, load_skill
 
-# Load all visualization skills
-_skills_text = load_all_skills()
+# Load visualization skills (excalidraw loaded separately — it targets MCP tools, not widgetRenderer)
+_widget_skills_text = load_all_skills(exclude=["excalidraw-diagram-skill"])
+_excalidraw_skill_text = load_skill("excalidraw-diagram-skill")
 
 agent = create_agent(
     model=ChatOpenAI(model="gpt-5.4-2026-03-05"),
@@ -46,7 +47,18 @@
 
         Follow the skills below for how to produce high-quality visuals:
 
-        {_skills_text}
+        {_widget_skills_text}
+
+        ## Excalidraw Diagramming Skills
+
+        You also have access to Excalidraw MCP tools (`Excalidraw:read_me` and
+        `Excalidraw:create_view`) for creating animated, interactive diagrams.
+
+        When a user asks you to draw a diagram, flowchart, architecture map, or any
+        visual that would benefit from the Excalidraw canvas — use the Excalidraw MCP
+        tools instead of `widgetRenderer`. Follow the skill below exactly:
+
+        {_excalidraw_skill_text}
     """,
 )
 

apps/agent/skills/__init__.py

@@ -17,10 +17,17 @@ def load_skill(name: str) -> str:
     return path.read_text()
 
 
-def load_all_skills() -> str:
-    """Load and concatenate all .txt skill files in this directory."""
+def load_all_skills(exclude: list[str] | None = None) -> str:
+    """Load and concatenate all .txt skill files in this directory.
+
+    Args:
+        exclude: Optional list of skill names (without extension) to skip.
+    """
+    skip = set(exclude or [])
     parts: list[str] = []
     for path in sorted(_SKILLS_DIR.glob("*.txt")):
+        if path.stem in skip:
+            continue
         parts.append(f"\n\n{'='*60}\n# SKILL: {path.stem}\n{'='*60}\n\n")
         parts.append(path.read_text())
     return "".join(parts)