feat(studio): add email-security-analyst sample; agents list UX

web/packages/studio/AGENTS.md

@@ -54,3 +54,44 @@ Wrong — transparent dropdown:
 
 Exception: call sites that fill SelectContent with custom children containing their
 own background (e.g. a sticky Block with bg-surface) are fine as-is.
+
+### KUI form gotchas — the silent-failure set
+
+Each compiles, lints and typechecks clean, then renders nothing wrong — the element is
+simply inert or invisible.
+
+**`slotError` needs `status="error"` to render.** Without a status `FormField` shows
+`slotHelp` instead and drops the message. `Controlled*` wrappers set status from
+react-hook-form only, so any other error source must set it too (`formFieldProps` spreads
+last, so it wins):
+
+```tsx
+formFieldProps={{ slotError: fieldError, status: fieldError ? 'error' : undefined }}
+```
+
+**`FormModal.disabled` blocks closing.** It means "busy": it kills Cancel and stops
+`handleUserClose`, trapping the user. For validation use `submitDisabled`.
+
+```tsx
+disabled={isPending}         // busy → intentionally locked
+submitDisabled={!isValid}    // invalid → submit blocked, dismiss still works
+```
+
+**`Text` has no `color="danger"`, and `text-danger` is not an emitted utility.** `danger`
+exists only on `Button`; the token is `--text-color-feedback-danger`. Devtools showing
+**"Inherited from"** means no rule matched at all, not that yours lost a specificity fight.
+
+```tsx
+<Text className="text-[var(--text-color-feedback-danger)]">…</Text>
+```
+
+**Modals seeded from a prop must re-seed on `open`.** `useForm({ defaultValues })` reads
+once at mount; a persistently-rendered modal mounts before the prop exists and keeps the
+empty default forever. Callers passing a constant work by accident, so the bug arrives
+with the second caller.
+
+```tsx
+useEffect(() => {
+  resetForm(makeDefaultValues(agentProp)); // NOT `if (!open)` — that only resets on close
+}, [open, agentProp, resetForm]);
+```

web/packages/studio/public/sample-agents/email-phishing-analyzer/agent.yml

@@ -18,7 +18,7 @@ llms:
   llm:
     _type: openai
     api_key: not-used
-    model_name: ${NEMO_DEFAULT_MODEL}
+    model_name: default/nvidia-nemotron-3-nano-30b-a3b
     temperature: 0.0
     max_tokens: 1024
 workflow:

web/packages/studio/public/sample-agents/email-phishing-analyzer/eval-config.json

@@ -1,199 +1,60 @@
 {
-  "tasks": [
+  "dataset": "<workspace>/<fileset>#dataset.jsonl",
+  "prompt_template": "Subject: {{ item.subject }}\nFrom: {{ item.sender }}\n\n{{ item.body }}",
+  "metrics": [
     {
-      "id": "email-0",
-      "intent": "Classify this email as phishing or benign",
-      "inputs": {
-        "instruction": "Subject: Claim Your Free iPhone Now!\nFrom: prize@example.com\n\nDear valued customer,\nCongratulations! You have been selected to receive a brand new iPhone absolutely free. To claim your prize, simply click the link below and provide your shipping address.\nhttp://malicious-link.example.com/claim\nThis offer is limited, so act fast!"
-      },
-      "reference": {
-        "label": "phishing"
-      },
-      "metadata": [
-        {
-          "key": "arrival_time",
-          "value": "2023-05-14 10:15:30"
-        },
-        {
-          "key": "intents",
-          "value": "{'money': {'label': 'Money', 'id': 0, 'score': 0.9998}, 'banking': {'label': 'Personal', 'id': 1, 'score': 0.9997}, 'crypto': {'label': 'NonCrypto', 'id': 1, 'score': 0.9996}}"
-        },
-        {
-          "key": "source",
-          "value": "gift"
-        },
-        {
-          "key": "extra_info",
-          "value": "unverified"
-        }
-      ]
-    },
-    {
-      "id": "email-1",
-      "intent": "Classify this email as phishing or benign",
-      "inputs": {
-        "instruction": "Subject: Urgent: Your Account Has Been Suspended\nFrom: security-alerts@bank.com\n\nHello,\nWe have detected unusual activity on your account. To prevent suspension, please verify your identity by clicking the link below and entering your credentials.\nhttp://verify-account.example.com\nIf you do not verify within 24 hours, your account will be disabled.\nThank you,\nSupport Team"
-      },
-      "reference": {
-        "label": "phishing"
-      },
-      "metadata": [
-        {
-          "key": "arrival_time",
-          "value": "2023-06-22 14:07:12"
-        },
-        {
-          "key": "intents",
-          "value": "{'money': {'label': 'Money', 'id': 0, 'score': 0.9999}, 'banking': {'label': 'Personal', 'id': 1, 'score': 0.9999}, 'crypto': {'label': 'NonCrypto', 'id': 1, 'score': 0.9995}}"
-        },
-        {
-          "key": "source",
-          "value": "password"
-        },
-        {
-          "key": "extra_info",
-          "value": "suspicious"
-        }
-      ]
-    },
-    {
-      "id": "email-2",
-      "intent": "Classify this email as phishing or benign",
-      "inputs": {
-        "instruction": "Subject: Important: Invoice Attached\nFrom: accounts@shop-example.com\n\nHi there,\nPlease find the invoice attached for your recent purchase. Click here to view the details.\nhttp://invoice-example.com/view?invoice=12345\nIf you have any questions, feel free to contact us.\nBest regards,\nCustomer Service"
-      },
-      "reference": {
-        "label": "phishing"
-      },
-      "metadata": [
-        {
-          "key": "arrival_time",
-          "value": "2023-07-01 09:30:45"
-        },
-        {
-          "key": "intents",
-          "value": "{'money': {'label': 'Money', 'id': 0, 'score': 0.9997}, 'banking': {'label': 'Personal', 'id': 1, 'score': 0.9998}, 'crypto': {'label': 'NonCrypto', 'id': 1, 'score': 0.9994}}"
-        },
-        {
-          "key": "source",
-          "value": "money"
-        },
-        {
-          "key": "extra_info",
-          "value": "pending"
-        }
-      ]
-    },
-    {
-      "id": "email-3",
-      "intent": "Classify this email as phishing or benign",
-      "inputs": {
-        "instruction": "Subject: Project Meeting Reminder\nFrom: bob@example.com\n\nHi Team,\nJust wanted to remind you about our project update meeting on Friday at 2pm. Please let me know if you can attend.\nThanks!\n-Bob"
-      },
-      "reference": {
-        "label": "benign"
-      },
-      "metadata": [
-        {
-          "key": "arrival_time",
-          "value": "2023-08-10 15:30:00"
-        },
-        {
-          "key": "intents",
-          "value": "{'money': {'label': 'NonMoney', 'id': 1, 'score': 0.9995}, 'banking': {'label': 'NonPersonal', 'id': 1, 'score': 0.9995}, 'crypto': {'label': 'NonCrypto', 'id': 1, 'score': 0.9994}}"
-        },
-        {
-          "key": "source",
-          "value": "meeting"
-        },
-        {
-          "key": "extra_info",
-          "value": "trusted"
-        }
-      ]
-    },
-    {
-      "id": "email-4",
-      "intent": "Classify this email as phishing or benign",
-      "inputs": {
-        "instruction": "Subject: Invoice Follow-up\nFrom: alice@company.com\n\nHi John,\nPlease find the invoice #1234 attached for your recent purchase. Let me know if you have any questions.\nBest regards,\nAlice"
-      },
-      "reference": {
-        "label": "benign"
-      },
-      "metadata": [
-        {
-          "key": "arrival_time",
-          "value": "2023-09-01 09:15:22"
-        },
-        {
-          "key": "intents",
-          "value": "{'money': {'label': 'Money', 'id': 0, 'score': 0.9996}, 'banking': {'label': 'Personal', 'id': 1, 'score': 0.9996}, 'crypto': {'label': 'NonCrypto', 'id': 1, 'score': 0.9993}}"
-        },
-        {
-          "key": "source",
-          "value": "finance"
-        },
-        {
-          "key": "extra_info",
-          "value": "trusted"
-        }
-      ]
-    }
-  ],
-  "metric": {
-    "bundle_kind": "metric-bundle",
-    "bundle_format_version": "v1",
-    "metric_type": "llm-judge",
-    "metadata": {
-      "description": null,
-      "labels": {}
-    },
-    "outputs": [
-      {
-        "name": "accuracy",
-        "description": null,
-        "value_json_schema": {
-          "description": "Continuous numeric metric value.",
-          "title": "ContinuousScore",
-          "type": "number"
+      "bundle_kind": "metric-bundle",
+      "bundle_format_version": "v1",
+      "metric_type": "llm-judge",
+      "metadata": {
+        "description": "Verdict accuracy vs the dataset's label column.",
+        "labels": {}
+      },
+      "outputs": [
+        {
+          "name": "accuracy",
+          "description": null,
+          "value_json_schema": {
+            "description": "Continuous numeric metric value.",
+            "title": "ContinuousScore",
+            "type": "number"
+          }
         }
-      }
-    ],
-    "secrets": {},
-    "payload": {
-      "kind": "inline",
-      "metric": {
-        "type": "llm-judge",
-        "model": "default/nvidia-nemotron-3-super-120b-a12b",
-        "prompt_template": {
-          "messages": [
+      ],
+      "secrets": {},
+      "payload": {
+        "kind": "inline",
+        "metric": {
+          "type": "llm-judge",
+          "model": "default/nvidia-nemotron-3-super-120b-a12b",
+          "prompt_template": {
+            "messages": [
+              {
+                "role": "user",
+                "content": "You are scoring an email phishing classifier. The expected answer is the ground-truth label: \"{{ item.label }}\". The agent's response may contain reasoning; extract ONLY its final verdict (the last clear \"phishing\" or \"benign\").\n\nAgent response:\n{{ sample.output_text }}\n\nRespond with a JSON object {\"accuracy\": <value>} where accuracy is 1 if the extracted verdict matches the expected label, else 0."
+              }
+            ]
+          },
+          "scores": [
             {
-              "role": "user",
-              "content": "You are scoring an email phishing classifier. The expected answer is the ground-truth label: \"{{ item.reference.label }}\". The agent's response may contain reasoning; extract ONLY its final verdict (the last clear \"phishing\" or \"benign\").\n\nAgent response:\n{{ sample.output_text }}\n\nRespond with a JSON object {\"accuracy\": <value>} where accuracy is 1 if the extracted verdict matches the expected label, else 0."
+              "name": "accuracy",
+              "minimum": 0,
+              "maximum": 1
             }
-          ]
-        },
-        "scores": [
-          {
-            "name": "accuracy",
-            "minimum": 0,
-            "maximum": 1
-          }
-        ],
-        "inference": {
-          "max_tokens": 1024,
-          "extra_body": {
-            "nvext": {
-              "max_thinking_tokens": 256
+          ],
+          "inference": {
+            "max_tokens": 1024,
+            "extra_body": {
+              "nvext": {
+                "max_thinking_tokens": 256
+              }
             }
+          },
+          "reasoning": {
+            "end_token": "</think>"
           }
-        },
-        "reasoning": {
-          "end_token": "</think>"
         }
       }
     }
-  },
-  "max_concurrent_tasks": 1
+  ]
 }