Fix skill execution: wire run_skill to actually invoke agent, show results in modal

dashboard/pages/skills.js

@@ -156,6 +156,7 @@ async function quickRunSkill(name) {
         <option value="gemini">Gemini CLI</option>
       </select>
     </div>
+    <div id="skillResult" style="display:none"></div>
   `, `
     <button class="btn btn-ghost" onclick="closeModal()">Cancel</button>
     <button class="btn btn-primary" onclick="executeSkillRun('${name}')">▶ Run</button>
@@ -165,11 +166,32 @@ async function quickRunSkill(name) {
 async function executeSkillRun(name) {
   const input = document.getElementById('qrsInput').value;
   const agent = document.getElementById('qrsAgent').value;
+  const runBtn = document.querySelector('#modalContainer .btn-primary');
+  const resultArea = document.getElementById('skillResult');
+
+  if (runBtn) { runBtn.disabled = true; runBtn.textContent = '⏳ Running...'; }
+  if (resultArea) {
+    resultArea.style.display = 'block';
+    resultArea.innerHTML = '<div class="loading" style="padding:20px"><div class="loading-spinner"></div><span style="margin-left:8px">Executing skill...</span></div>';
+  }
+
   try {
     const r = await api.runSkill(name, input, agent);
-    closeModal();
-    showToast(`"${name}" dispatched to ${r.agent} #${r.run_id}`, 'success');
+    if (resultArea) {
+      const outputText = r.output || '(no output)';
+      resultArea.innerHTML = `
+        <div class="card" style="margin-top:8px">
+          <div class="card-header" style="border-color:var(--green-dim)">
+            <span class="card-title" style="color:var(--green)">✓ Completed — ${r.agent} #${r.run_id}</span>
+          </div>
+          <pre style="max-height:400px;overflow:auto;font-size:12px;white-space:pre-wrap;margin:0;padding:12px;background:var(--bg-code, #1a1a2e);border-radius:0 0 8px 8px">${escapeHtml(outputText)}</pre>
+        </div>`;
+    }
+    if (runBtn) { runBtn.textContent = '✓ Done'; runBtn.disabled = false; }
   } catch (err) {
-    showToast(`Error: ${err.message}`, 'error');
+    if (resultArea) {
+      resultArea.innerHTML = `<div class="empty-state" style="padding:20px"><div class="empty-state-icon">⚠</div><div class="empty-state-title">Error</div><div class="empty-state-desc">${escapeHtml(err.message)}</div></div>`;
+    }
+    if (runBtn) { runBtn.textContent = '▶ Run'; runBtn.disabled = false; }
   }
 }

server.py

@@ -205,6 +205,10 @@ def run_skill(name: str, req: Optional[SkillRunRequest] = None):
     agent_choice = req.agent if req else "auto"
     skill_input = req.input if req else ""
 
+    # Read skill files
+    skill_md = read_file(path / "SKILL.md")
+    learnings = read_file(path / "learnings.md")
+
     # Determine which agent based on skill type
     if agent_choice == "auto":
         devops_keywords = ["devops", "audit", "deploy", "k8s", "gcp", "infra", "terraform"]
@@ -214,23 +218,65 @@ def run_skill(name: str, req: Optional[SkillRunRequest] = None):
         elif any(k in name for k in research_keywords):
             agent_choice = "gemini"
         else:
-            agent_choice = "opencode"
+            # Check SKILL.md for explicit agent assignment
+            for line in skill_md.split('\n'):
+                line = line.strip()
+                if "Primary:" in line:
+                    candidate = line.split(":")[-1].strip().lower()
+                    if candidate in ("opencode", "hermes", "gemini"):
+                        agent_choice = candidate
+                        break
+            if agent_choice == "auto":
+                agent_choice = "opencode"
+
+    # Build prompt from skill instructions + learnings + user input
+    prompt = f"Execute the '{name}' skill.\n\n"
+    if skill_md:
+        prompt += f"## Skill Instructions\n{skill_md}\n\n"
+    if learnings and learnings.strip():
+        prompt += f"## Past Learnings\n{learnings}\n\n"
+    if skill_input:
+        prompt += f"## User Input\n{skill_input}"
+
+    run_id = str(uuid.uuid4())[:8]
+
+    # Execute via agent
+    try:
+        response_text = execute_agent(agent_choice, prompt)
+    except subprocess.TimeoutExpired:
+        response_text = f"⏱ Skill '{name}' timed out on agent '{agent_choice}'."
+    except FileNotFoundError:
+        response_text = f"⚠ Agent '{agent_choice}' CLI not installed. Install it and try again."
+    except Exception as e:
+        response_text = f"⚠ Error executing skill: {str(e)}"
+
+    # Save output to learnings.md
+    timestamp = get_timestamp()[:10]
+    existing = read_file(path / "learnings.md")
+    new_entry = (
+        f"\n## {timestamp} (Run {run_id})\n"
+        f"- Agent: {agent_choice}\n"
+        f"- Input: {skill_input or '(none)'}\n"
+        f"- Output: {response_text[:500]}\n"
+    )
+    write_file(path / "learnings.md", existing + new_entry)
 
     # Log execution
-    run_id = str(uuid.uuid4())[:8]
     append_audit({
         "action": "skill_run",
         "skill": name,
         "agent": agent_choice,
         "run_id": run_id,
+        "output_preview": response_text[:100],
     })
 
     return {
-        "status": "started",
+        "status": "completed",
         "run_id": run_id,
         "skill": name,
         "agent": agent_choice,
-        "message": f"Skill '{name}' dispatched to {agent_choice}",
+        "output": response_text,
+        "message": f"Skill '{name}' completed via {agent_choice}",
     }
 
 @app.get("/api/skills/{name}/eval")