@yanlinglabs/winter-conformance 0.0.14 → 0.0.16

package/goldens/p6-openai-responses-fake.trace.json

@@ -57,7 +57,13 @@
       ],
       "output_style": "default",
       "skills": [],
-      "plugins": []
+      "plugins": [],
+      "agents": [
+        "claude",
+        "Explore",
+        "general-purpose",
+        "Plan"
+      ]
     }
   },
   {
@@ -95,7 +101,8 @@
           {
             "type": "tool_result",
             "tool_use_id": "call_p6",
-            "content": "Error: path not found: /winter-fixture"
+            "content": "Error: path not found: /winter-fixture",
+            "is_error": true
           }
         ]
       }

package/goldens/p6-resolution-failure.trace.json

@@ -48,7 +48,13 @@
       ],
       "output_style": "default",
       "skills": [],
-      "plugins": []
+      "plugins": [],
+      "agents": [
+        "claude",
+        "Explore",
+        "general-purpose",
+        "Plan"
+      ]
     }
   },
   {

package/goldens/plain-query.trace.json

@@ -48,7 +48,13 @@
       ],
       "output_style": "default",
       "skills": [],
-      "plugins": []
+      "plugins": [],
+      "agents": [
+        "claude",
+        "Explore",
+        "general-purpose",
+        "Plan"
+      ]
     }
   },
   {
@@ -61,7 +67,7 @@
         "content": [
           {
             "type": "text",
-            "text": "echo: <system-reminder>\nAuto-memory (injected by the runtime, not typed by the user):\nAuto-memory for this project lives at /winter-home/projects/-winter-fixture/memory. The directory is created on demand and there are no memory tools: read and write it with the ordinary file tools, exactly like any other directory.\n\n/winter-home/projects/-winter-fixture/memory/MEMORY.md is the INDEX, and only its first 200 lines / 25 KB are loaded into a session. Keep every index entry to one line; the substance belongs in the topic file it points at, which you can read on demand when it turns out to matter.\n\nTo record something worth having in a LATER session -- a standing preference, a correction you were given, a durable constraint of this project -- write it to its own `<slug>.md` in that directory and add a one-line pointer to the index: `- [<slug>](<slug>.md) — <one-line summary>`.\n\nRevise a fact by rewriting its file and its index line, never by adding a near-duplicate under a new name; remove both once it stops being true. An index full of stale near-duplicates is worse than an empty one, because it costs the same and misleads.\n\nDo not record what the repository already records. Code, configuration, documentation and WINTER.md are durable on their own; memory is for what is true about this project or this user and lives nowhere in the tree.\n</system-reminder>\n\nhi"
+            "text": "echo: <system-reminder>\nAvailable agent types for the Agent tool:\n- claude: Catch-all for any task that doesn't fit a more specific agent. (Tools: *)\n- Explore: Fast read-only search agent for locating code. Use it to find files by pattern (eg. \"src/components/**/*.tsx\"), grep for symbols or keywords (eg. \"API endpoints\"), or answer \"where is X defined / which files reference Y.\" Do NOT use it for code review, design-doc auditing, cross-file consistency checks, or open-ended analysis — it reads excerpts rather than whole files and will miss content past its read window. When calling, specify search breadth: \"quick\" for a single targeted lookup, \"medium\" for moderate exploration, or \"very thorough\" to search across multiple locations and naming conventions. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\n- general-purpose: General-purpose agent for researching complex questions, searching for code, and executing multi-step tasks. When you are searching for a keyword or file and are not confident that you will find the right match in the first few tries use this agent to perform the search for you. (Tools: *)\n- Plan: Software architect agent for designing implementation plans. Use this when you need to plan the implementation strategy for a task. Returns step-by-step plans, identifies critical files, and considers architectural trade-offs. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\n\nWhen you launch multiple agents for independent work, send them in a single message with multiple tool uses so they run concurrently.\n</system-reminder>\n\n<system-reminder>\nAs you answer the user's questions, you can use the following context:\n# currentDate\nToday's date is <FIXTURE-DATE>.\n\n      IMPORTANT: this context may or may not be relevant to your tasks. You should not respond to this context unless it is highly relevant to your task.\n</system-reminder>\n\n\nhi"
           }
         ]
       }
@@ -75,7 +81,7 @@
       "type": "result",
       "subtype": "success",
       "is_error": false,
-      "result": "echo: <system-reminder>\nAuto-memory (injected by the runtime, not typed by the user):\nAuto-memory for this project lives at /winter-home/projects/-winter-fixture/memory. The directory is created on demand and there are no memory tools: read and write it with the ordinary file tools, exactly like any other directory.\n\n/winter-home/projects/-winter-fixture/memory/MEMORY.md is the INDEX, and only its first 200 lines / 25 KB are loaded into a session. Keep every index entry to one line; the substance belongs in the topic file it points at, which you can read on demand when it turns out to matter.\n\nTo record something worth having in a LATER session -- a standing preference, a correction you were given, a durable constraint of this project -- write it to its own `<slug>.md` in that directory and add a one-line pointer to the index: `- [<slug>](<slug>.md) — <one-line summary>`.\n\nRevise a fact by rewriting its file and its index line, never by adding a near-duplicate under a new name; remove both once it stops being true. An index full of stale near-duplicates is worse than an empty one, because it costs the same and misleads.\n\nDo not record what the repository already records. Code, configuration, documentation and WINTER.md are durable on their own; memory is for what is true about this project or this user and lives nowhere in the tree.\n</system-reminder>\n\nhi",
+      "result": "echo: <system-reminder>\nAvailable agent types for the Agent tool:\n- claude: Catch-all for any task that doesn't fit a more specific agent. (Tools: *)\n- Explore: Fast read-only search agent for locating code. Use it to find files by pattern (eg. \"src/components/**/*.tsx\"), grep for symbols or keywords (eg. \"API endpoints\"), or answer \"where is X defined / which files reference Y.\" Do NOT use it for code review, design-doc auditing, cross-file consistency checks, or open-ended analysis — it reads excerpts rather than whole files and will miss content past its read window. When calling, specify search breadth: \"quick\" for a single targeted lookup, \"medium\" for moderate exploration, or \"very thorough\" to search across multiple locations and naming conventions. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\n- general-purpose: General-purpose agent for researching complex questions, searching for code, and executing multi-step tasks. When you are searching for a keyword or file and are not confident that you will find the right match in the first few tries use this agent to perform the search for you. (Tools: *)\n- Plan: Software architect agent for designing implementation plans. Use this when you need to plan the implementation strategy for a task. Returns step-by-step plans, identifies critical files, and considers architectural trade-offs. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\n\nWhen you launch multiple agents for independent work, send them in a single message with multiple tool uses so they run concurrently.\n</system-reminder>\n\n<system-reminder>\nAs you answer the user's questions, you can use the following context:\n# currentDate\nToday's date is <FIXTURE-DATE>.\n\n      IMPORTANT: this context may or may not be relevant to your tasks. You should not respond to this context unless it is highly relevant to your task.\n</system-reminder>\n\n\nhi",
       "permission_denials": []
     }
   }

package/goldens/resume.trace.json

@@ -48,7 +48,13 @@
       ],
       "output_style": "default",
       "skills": [],
-      "plugins": []
+      "plugins": [],
+      "agents": [
+        "claude",
+        "Explore",
+        "general-purpose",
+        "Plan"
+      ]
     }
   },
   {
@@ -61,7 +67,7 @@
         "content": [
           {
             "type": "text",
-            "text": "[{\"role\":\"user\",\"content\":\"<system-reminder>\\nAuto-memory (injected by the runtime, not typed by the user):\\nAuto-memory for this project lives at /winter-home/projects/-winter-fixture/memory. The directory is created on demand and there are no memory tools: read and write it with the ordinary file tools, exactly like any other directory.\\n\\n/winter-home/projects/-winter-fixture/memory/MEMORY.md is the INDEX, and only its first 200 lines / 25 KB are loaded into a session. Keep every index entry to one line; the substance belongs in the topic file it points at, which you can read on demand when it turns out to matter.\\n\\nTo record something worth having in a LATER session -- a standing preference, a correction you were given, a durable constraint of this project -- write it to its own `<slug>.md` in that directory and add a one-line pointer to the index: `- [<slug>](<slug>.md) — <one-line summary>`.\\n\\nRevise a fact by rewriting its file and its index line, never by adding a near-duplicate under a new name; remove both once it stops being true. An index full of stale near-duplicates is worse than an empty one, because it costs the same and misleads.\\n\\nDo not record what the repository already records. Code, configuration, documentation and WINTER.md are durable on their own; memory is for what is true about this project or this user and lives nowhere in the tree.\\n</system-reminder>\\n\\nfirst\"}]"
+            "text": "[{\"role\":\"user\",\"content\":[{\"type\":\"text\",\"text\":\"<system-reminder>\\nAvailable agent types for the Agent tool:\\n- claude: Catch-all for any task that doesn't fit a more specific agent. (Tools: *)\\n- Explore: Fast read-only search agent for locating code. Use it to find files by pattern (eg. \\\"src/components/**/*.tsx\\\"), grep for symbols or keywords (eg. \\\"API endpoints\\\"), or answer \\\"where is X defined / which files reference Y.\\\" Do NOT use it for code review, design-doc auditing, cross-file consistency checks, or open-ended analysis — it reads excerpts rather than whole files and will miss content past its read window. When calling, specify search breadth: \\\"quick\\\" for a single targeted lookup, \\\"medium\\\" for moderate exploration, or \\\"very thorough\\\" to search across multiple locations and naming conventions. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\n- general-purpose: General-purpose agent for researching complex questions, searching for code, and executing multi-step tasks. When you are searching for a keyword or file and are not confident that you will find the right match in the first few tries use this agent to perform the search for you. (Tools: *)\\n- Plan: Software architect agent for designing implementation plans. Use this when you need to plan the implementation strategy for a task. Returns step-by-step plans, identifies critical files, and considers architectural trade-offs. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\n\\nWhen you launch multiple agents for independent work, send them in a single message with multiple tool uses so they run concurrently.\\n</system-reminder>\\n\"},{\"type\":\"text\",\"text\":\"<system-reminder>\\nAs you answer the user's questions, you can use the following context:\\n# currentDate\\nToday's date is <FIXTURE-DATE>.\\n\\n      IMPORTANT: this context may or may not be relevant to your tasks. You should not respond to this context unless it is highly relevant to your task.\\n</system-reminder>\\n\\n\"},{\"type\":\"text\",\"text\":\"first\"}]}]"
           }
         ]
       }
@@ -75,7 +81,7 @@
       "type": "result",
       "subtype": "success",
       "is_error": false,
-      "result": "[{\"role\":\"user\",\"content\":\"<system-reminder>\\nAuto-memory (injected by the runtime, not typed by the user):\\nAuto-memory for this project lives at /winter-home/projects/-winter-fixture/memory. The directory is created on demand and there are no memory tools: read and write it with the ordinary file tools, exactly like any other directory.\\n\\n/winter-home/projects/-winter-fixture/memory/MEMORY.md is the INDEX, and only its first 200 lines / 25 KB are loaded into a session. Keep every index entry to one line; the substance belongs in the topic file it points at, which you can read on demand when it turns out to matter.\\n\\nTo record something worth having in a LATER session -- a standing preference, a correction you were given, a durable constraint of this project -- write it to its own `<slug>.md` in that directory and add a one-line pointer to the index: `- [<slug>](<slug>.md) — <one-line summary>`.\\n\\nRevise a fact by rewriting its file and its index line, never by adding a near-duplicate under a new name; remove both once it stops being true. An index full of stale near-duplicates is worse than an empty one, because it costs the same and misleads.\\n\\nDo not record what the repository already records. Code, configuration, documentation and WINTER.md are durable on their own; memory is for what is true about this project or this user and lives nowhere in the tree.\\n</system-reminder>\\n\\nfirst\"}]",
+      "result": "[{\"role\":\"user\",\"content\":[{\"type\":\"text\",\"text\":\"<system-reminder>\\nAvailable agent types for the Agent tool:\\n- claude: Catch-all for any task that doesn't fit a more specific agent. (Tools: *)\\n- Explore: Fast read-only search agent for locating code. Use it to find files by pattern (eg. \\\"src/components/**/*.tsx\\\"), grep for symbols or keywords (eg. \\\"API endpoints\\\"), or answer \\\"where is X defined / which files reference Y.\\\" Do NOT use it for code review, design-doc auditing, cross-file consistency checks, or open-ended analysis — it reads excerpts rather than whole files and will miss content past its read window. When calling, specify search breadth: \\\"quick\\\" for a single targeted lookup, \\\"medium\\\" for moderate exploration, or \\\"very thorough\\\" to search across multiple locations and naming conventions. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\n- general-purpose: General-purpose agent for researching complex questions, searching for code, and executing multi-step tasks. When you are searching for a keyword or file and are not confident that you will find the right match in the first few tries use this agent to perform the search for you. (Tools: *)\\n- Plan: Software architect agent for designing implementation plans. Use this when you need to plan the implementation strategy for a task. Returns step-by-step plans, identifies critical files, and considers architectural trade-offs. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\n\\nWhen you launch multiple agents for independent work, send them in a single message with multiple tool uses so they run concurrently.\\n</system-reminder>\\n\"},{\"type\":\"text\",\"text\":\"<system-reminder>\\nAs you answer the user's questions, you can use the following context:\\n# currentDate\\nToday's date is <FIXTURE-DATE>.\\n\\n      IMPORTANT: this context may or may not be relevant to your tasks. You should not respond to this context unless it is highly relevant to your task.\\n</system-reminder>\\n\\n\"},{\"type\":\"text\",\"text\":\"first\"}]}]",
       "permission_denials": []
     }
   },
@@ -128,7 +134,13 @@
       ],
       "output_style": "default",
       "skills": [],
-      "plugins": []
+      "plugins": [],
+      "agents": [
+        "claude",
+        "Explore",
+        "general-purpose",
+        "Plan"
+      ]
     }
   },
   {
@@ -141,7 +153,7 @@
         "content": [
           {
             "type": "text",
-            "text": "[{\"role\":\"user\",\"content\":\"first\"},{\"role\":\"assistant\",\"content\":\"[{\\\"role\\\":\\\"user\\\",\\\"content\\\":\\\"<system-reminder>\\\\nAuto-memory (injected by the runtime, not typed by the user):\\\\nAuto-memory for this project lives at /winter-home/projects/-winter-fixture/memory. The directory is created on demand and there are no memory tools: read and write it with the ordinary file tools, exactly like any other directory.\\\\n\\\\n/winter-home/projects/-winter-fixture/memory/MEMORY.md is the INDEX, and only its first 200 lines / 25 KB are loaded into a session. Keep every index entry to one line; the substance belongs in the topic file it points at, which you can read on demand when it turns out to matter.\\\\n\\\\nTo record something worth having in a LATER session -- a standing preference, a correction you were given, a durable constraint of this project -- write it to its own `<slug>.md` in that directory and add a one-line pointer to the index: `- [<slug>](<slug>.md) — <one-line summary>`.\\\\n\\\\nRevise a fact by rewriting its file and its index line, never by adding a near-duplicate under a new name; remove both once it stops being true. An index full of stale near-duplicates is worse than an empty one, because it costs the same and misleads.\\\\n\\\\nDo not record what the repository already records. Code, configuration, documentation and WINTER.md are durable on their own; memory is for what is true about this project or this user and lives nowhere in the tree.\\\\n</system-reminder>\\\\n\\\\nfirst\\\"}]\"},{\"role\":\"user\",\"content\":\"<system-reminder>\\nAuto-memory (injected by the runtime, not typed by the user):\\nAuto-memory for this project lives at /winter-home/projects/-winter-fixture/memory. The directory is created on demand and there are no memory tools: read and write it with the ordinary file tools, exactly like any other directory.\\n\\n/winter-home/projects/-winter-fixture/memory/MEMORY.md is the INDEX, and only its first 200 lines / 25 KB are loaded into a session. Keep every index entry to one line; the substance belongs in the topic file it points at, which you can read on demand when it turns out to matter.\\n\\nTo record something worth having in a LATER session -- a standing preference, a correction you were given, a durable constraint of this project -- write it to its own `<slug>.md` in that directory and add a one-line pointer to the index: `- [<slug>](<slug>.md) — <one-line summary>`.\\n\\nRevise a fact by rewriting its file and its index line, never by adding a near-duplicate under a new name; remove both once it stops being true. An index full of stale near-duplicates is worse than an empty one, because it costs the same and misleads.\\n\\nDo not record what the repository already records. Code, configuration, documentation and WINTER.md are durable on their own; memory is for what is true about this project or this user and lives nowhere in the tree.\\n</system-reminder>\\n\\nsecond\"}]"
+            "text": "[{\"role\":\"user\",\"content\":[{\"type\":\"text\",\"text\":\"<system-reminder>\\nAvailable agent types for the Agent tool:\\n- claude: Catch-all for any task that doesn't fit a more specific agent. (Tools: *)\\n- Explore: Fast read-only search agent for locating code. Use it to find files by pattern (eg. \\\"src/components/**/*.tsx\\\"), grep for symbols or keywords (eg. \\\"API endpoints\\\"), or answer \\\"where is X defined / which files reference Y.\\\" Do NOT use it for code review, design-doc auditing, cross-file consistency checks, or open-ended analysis — it reads excerpts rather than whole files and will miss content past its read window. When calling, specify search breadth: \\\"quick\\\" for a single targeted lookup, \\\"medium\\\" for moderate exploration, or \\\"very thorough\\\" to search across multiple locations and naming conventions. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\n- general-purpose: General-purpose agent for researching complex questions, searching for code, and executing multi-step tasks. When you are searching for a keyword or file and are not confident that you will find the right match in the first few tries use this agent to perform the search for you. (Tools: *)\\n- Plan: Software architect agent for designing implementation plans. Use this when you need to plan the implementation strategy for a task. Returns step-by-step plans, identifies critical files, and considers architectural trade-offs. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\n\\nWhen you launch multiple agents for independent work, send them in a single message with multiple tool uses so they run concurrently.\\n</system-reminder>\\n\"},{\"type\":\"text\",\"text\":\"<system-reminder>\\nAs you answer the user's questions, you can use the following context:\\n# currentDate\\nToday's date is <FIXTURE-DATE>.\\n\\n      IMPORTANT: this context may or may not be relevant to your tasks. You should not respond to this context unless it is highly relevant to your task.\\n</system-reminder>\\n\\n\"},{\"type\":\"text\",\"text\":\"first\"}]},{\"role\":\"assistant\",\"content\":\"[{\\\"role\\\":\\\"user\\\",\\\"content\\\":[{\\\"type\\\":\\\"text\\\",\\\"text\\\":\\\"<system-reminder>\\\\nAvailable agent types for the Agent tool:\\\\n- claude: Catch-all for any task that doesn't fit a more specific agent. (Tools: *)\\\\n- Explore: Fast read-only search agent for locating code. Use it to find files by pattern (eg. \\\\\\\"src/components/**/*.tsx\\\\\\\"), grep for symbols or keywords (eg. \\\\\\\"API endpoints\\\\\\\"), or answer \\\\\\\"where is X defined / which files reference Y.\\\\\\\" Do NOT use it for code review, design-doc auditing, cross-file consistency checks, or open-ended analysis — it reads excerpts rather than whole files and will miss content past its read window. When calling, specify search breadth: \\\\\\\"quick\\\\\\\" for a single targeted lookup, \\\\\\\"medium\\\\\\\" for moderate exploration, or \\\\\\\"very thorough\\\\\\\" to search across multiple locations and naming conventions. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\\\n- general-purpose: General-purpose agent for researching complex questions, searching for code, and executing multi-step tasks. When you are searching for a keyword or file and are not confident that you will find the right match in the first few tries use this agent to perform the search for you. (Tools: *)\\\\n- Plan: Software architect agent for designing implementation plans. Use this when you need to plan the implementation strategy for a task. Returns step-by-step plans, identifies critical files, and considers architectural trade-offs. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\\\n\\\\nWhen you launch multiple agents for independent work, send them in a single message with multiple tool uses so they run concurrently.\\\\n</system-reminder>\\\\n\\\"},{\\\"type\\\":\\\"text\\\",\\\"text\\\":\\\"<system-reminder>\\\\nAs you answer the user's questions, you can use the following context:\\\\n# currentDate\\\\nToday's date is <FIXTURE-DATE>.\\\\n\\\\n      IMPORTANT: this context may or may not be relevant to your tasks. You should not respond to this context unless it is highly relevant to your task.\\\\n</system-reminder>\\\\n\\\\n\\\"},{\\\"type\\\":\\\"text\\\",\\\"text\\\":\\\"first\\\"}]}]\"},{\"role\":\"user\",\"content\":\"second\"}]"
           }
         ]
       }
@@ -155,7 +167,7 @@
       "type": "result",
       "subtype": "success",
       "is_error": false,
-      "result": "[{\"role\":\"user\",\"content\":\"first\"},{\"role\":\"assistant\",\"content\":\"[{\\\"role\\\":\\\"user\\\",\\\"content\\\":\\\"<system-reminder>\\\\nAuto-memory (injected by the runtime, not typed by the user):\\\\nAuto-memory for this project lives at /winter-home/projects/-winter-fixture/memory. The directory is created on demand and there are no memory tools: read and write it with the ordinary file tools, exactly like any other directory.\\\\n\\\\n/winter-home/projects/-winter-fixture/memory/MEMORY.md is the INDEX, and only its first 200 lines / 25 KB are loaded into a session. Keep every index entry to one line; the substance belongs in the topic file it points at, which you can read on demand when it turns out to matter.\\\\n\\\\nTo record something worth having in a LATER session -- a standing preference, a correction you were given, a durable constraint of this project -- write it to its own `<slug>.md` in that directory and add a one-line pointer to the index: `- [<slug>](<slug>.md) — <one-line summary>`.\\\\n\\\\nRevise a fact by rewriting its file and its index line, never by adding a near-duplicate under a new name; remove both once it stops being true. An index full of stale near-duplicates is worse than an empty one, because it costs the same and misleads.\\\\n\\\\nDo not record what the repository already records. Code, configuration, documentation and WINTER.md are durable on their own; memory is for what is true about this project or this user and lives nowhere in the tree.\\\\n</system-reminder>\\\\n\\\\nfirst\\\"}]\"},{\"role\":\"user\",\"content\":\"<system-reminder>\\nAuto-memory (injected by the runtime, not typed by the user):\\nAuto-memory for this project lives at /winter-home/projects/-winter-fixture/memory. The directory is created on demand and there are no memory tools: read and write it with the ordinary file tools, exactly like any other directory.\\n\\n/winter-home/projects/-winter-fixture/memory/MEMORY.md is the INDEX, and only its first 200 lines / 25 KB are loaded into a session. Keep every index entry to one line; the substance belongs in the topic file it points at, which you can read on demand when it turns out to matter.\\n\\nTo record something worth having in a LATER session -- a standing preference, a correction you were given, a durable constraint of this project -- write it to its own `<slug>.md` in that directory and add a one-line pointer to the index: `- [<slug>](<slug>.md) — <one-line summary>`.\\n\\nRevise a fact by rewriting its file and its index line, never by adding a near-duplicate under a new name; remove both once it stops being true. An index full of stale near-duplicates is worse than an empty one, because it costs the same and misleads.\\n\\nDo not record what the repository already records. Code, configuration, documentation and WINTER.md are durable on their own; memory is for what is true about this project or this user and lives nowhere in the tree.\\n</system-reminder>\\n\\nsecond\"}]",
+      "result": "[{\"role\":\"user\",\"content\":[{\"type\":\"text\",\"text\":\"<system-reminder>\\nAvailable agent types for the Agent tool:\\n- claude: Catch-all for any task that doesn't fit a more specific agent. (Tools: *)\\n- Explore: Fast read-only search agent for locating code. Use it to find files by pattern (eg. \\\"src/components/**/*.tsx\\\"), grep for symbols or keywords (eg. \\\"API endpoints\\\"), or answer \\\"where is X defined / which files reference Y.\\\" Do NOT use it for code review, design-doc auditing, cross-file consistency checks, or open-ended analysis — it reads excerpts rather than whole files and will miss content past its read window. When calling, specify search breadth: \\\"quick\\\" for a single targeted lookup, \\\"medium\\\" for moderate exploration, or \\\"very thorough\\\" to search across multiple locations and naming conventions. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\n- general-purpose: General-purpose agent for researching complex questions, searching for code, and executing multi-step tasks. When you are searching for a keyword or file and are not confident that you will find the right match in the first few tries use this agent to perform the search for you. (Tools: *)\\n- Plan: Software architect agent for designing implementation plans. Use this when you need to plan the implementation strategy for a task. Returns step-by-step plans, identifies critical files, and considers architectural trade-offs. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\n\\nWhen you launch multiple agents for independent work, send them in a single message with multiple tool uses so they run concurrently.\\n</system-reminder>\\n\"},{\"type\":\"text\",\"text\":\"<system-reminder>\\nAs you answer the user's questions, you can use the following context:\\n# currentDate\\nToday's date is <FIXTURE-DATE>.\\n\\n      IMPORTANT: this context may or may not be relevant to your tasks. You should not respond to this context unless it is highly relevant to your task.\\n</system-reminder>\\n\\n\"},{\"type\":\"text\",\"text\":\"first\"}]},{\"role\":\"assistant\",\"content\":\"[{\\\"role\\\":\\\"user\\\",\\\"content\\\":[{\\\"type\\\":\\\"text\\\",\\\"text\\\":\\\"<system-reminder>\\\\nAvailable agent types for the Agent tool:\\\\n- claude: Catch-all for any task that doesn't fit a more specific agent. (Tools: *)\\\\n- Explore: Fast read-only search agent for locating code. Use it to find files by pattern (eg. \\\\\\\"src/components/**/*.tsx\\\\\\\"), grep for symbols or keywords (eg. \\\\\\\"API endpoints\\\\\\\"), or answer \\\\\\\"where is X defined / which files reference Y.\\\\\\\" Do NOT use it for code review, design-doc auditing, cross-file consistency checks, or open-ended analysis — it reads excerpts rather than whole files and will miss content past its read window. When calling, specify search breadth: \\\\\\\"quick\\\\\\\" for a single targeted lookup, \\\\\\\"medium\\\\\\\" for moderate exploration, or \\\\\\\"very thorough\\\\\\\" to search across multiple locations and naming conventions. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\\\n- general-purpose: General-purpose agent for researching complex questions, searching for code, and executing multi-step tasks. When you are searching for a keyword or file and are not confident that you will find the right match in the first few tries use this agent to perform the search for you. (Tools: *)\\\\n- Plan: Software architect agent for designing implementation plans. Use this when you need to plan the implementation strategy for a task. Returns step-by-step plans, identifies critical files, and considers architectural trade-offs. (Tools: All tools except Agent, Artifact, ExitPlanMode, Edit, Write, NotebookEdit)\\\\n\\\\nWhen you launch multiple agents for independent work, send them in a single message with multiple tool uses so they run concurrently.\\\\n</system-reminder>\\\\n\\\"},{\\\"type\\\":\\\"text\\\",\\\"text\\\":\\\"<system-reminder>\\\\nAs you answer the user's questions, you can use the following context:\\\\n# currentDate\\\\nToday's date is <FIXTURE-DATE>.\\\\n\\\\n      IMPORTANT: this context may or may not be relevant to your tasks. You should not respond to this context unless it is highly relevant to your task.\\\\n</system-reminder>\\\\n\\\\n\\\"},{\\\"type\\\":\\\"text\\\",\\\"text\\\":\\\"first\\\"}]}]\"},{\"role\":\"user\",\"content\":\"second\"}]",
       "permission_denials": []
     }
   }

package/goldens/sendmessage-child-round.trace.json

@@ -48,7 +48,13 @@
       ],
       "output_style": "default",
       "skills": [],
-      "plugins": []
+      "plugins": [],
+      "agents": [
+        "claude",
+        "Explore",
+        "general-purpose",
+        "Plan"
+      ]
     }
   },
   {
@@ -65,7 +71,8 @@
             "name": "Agent",
             "input": {
               "description": "message target",
-              "prompt": "child probe text"
+              "prompt": "child probe text",
+              "run_in_background": false
             }
           }
         ]
@@ -75,6 +82,55 @@
   {
     "sequence": 2,
     "direction": "runtime-to-host",
+    "kind": "system/task_started",
+    "payload": {
+      "type": "system",
+      "subtype": "task_started",
+      "task_id": "TASKID_1",
+      "tool_use_id": "agent-call-1",
+      "description": "message target",
+      "subagent_type": "general-purpose",
+      "is_backgrounded": false,
+      "spawn_depth": 1,
+      "task_type": "local_agent",
+      "prompt": "child probe text"
+    }
+  },
+  {
+    "sequence": 3,
+    "direction": "runtime-to-host",
+    "kind": "system/task_updated",
+    "payload": {
+      "type": "system",
+      "subtype": "task_updated",
+      "task_id": "TASKID_1",
+      "patch": {
+        "status": "completed",
+        "end_time": "<end_time>"
+      }
+    }
+  },
+  {
+    "sequence": 4,
+    "direction": "runtime-to-host",
+    "kind": "system/task_notification",
+    "payload": {
+      "type": "system",
+      "subtype": "task_notification",
+      "task_id": "TASKID_1",
+      "tool_use_id": "agent-call-1",
+      "status": "completed",
+      "output_file": "/winter-fixture-tasks/TASKID_1.output",
+      "summary": "child finished",
+      "usage": {
+        "total_tokens": "<total_tokens>",
+        "tool_uses": 0
+      }
+    }
+  },
+  {
+    "sequence": 5,
+    "direction": "runtime-to-host",
     "kind": "user",
     "payload": {
       "type": "user",
@@ -83,14 +139,14 @@
           {
             "type": "tool_result",
             "tool_use_id": "agent-call-1",
-            "content": "{\"agentId\":\"<scrubbed>\",\"content\":[{\"type\":\"text\",\"text\":\"child finished\"}],\"totalToolUseCount\":0,\"totalDurationMs\":\"<scrubbed>\",\"resolvedModel\":\"winter-test/echo\",\"prompt\":\"child probe text\"}"
+            "content": "{\"agentId\":\"<scrubbed>\",\"agentType\":\"general-purpose\",\"content\":[{\"type\":\"text\",\"text\":\"child finished\"}],\"totalToolUseCount\":0,\"totalDurationMs\":\"<scrubbed>\",\"resolvedModel\":\"winter-test/echo\",\"prompt\":\"child probe text\"}"
           }
         ]
       }
     }
   },
   {
-    "sequence": 3,
+    "sequence": 6,
     "direction": "runtime-to-host",
     "kind": "assistant",
     "payload": {
@@ -111,7 +167,7 @@
     }
   },
   {
-    "sequence": 4,
+    "sequence": 7,
     "direction": "runtime-to-host",
     "kind": "user",
     "payload": {
@@ -128,7 +184,7 @@
     }
   },
   {
-    "sequence": 5,
+    "sequence": 8,
     "direction": "runtime-to-host",
     "kind": "assistant",
     "payload": {
@@ -144,7 +200,7 @@
     }
   },
   {
-    "sequence": 6,
+    "sequence": 9,
     "direction": "runtime-to-host",
     "kind": "result",
     "payload": {

package/goldens/skill-invocation-round.trace.json

@@ -51,7 +51,13 @@
       "skills": [
         "p5probe"
       ],
-      "plugins": []
+      "plugins": [],
+      "agents": [
+        "claude",
+        "Explore",
+        "general-purpose",
+        "Plan"
+      ]
     }
   },
   {

package/goldens/structured-exhaustion-round.trace.json

@@ -49,7 +49,13 @@
       ],
       "output_style": "default",
       "skills": [],
-      "plugins": []
+      "plugins": [],
+      "agents": [
+        "claude",
+        "Explore",
+        "general-purpose",
+        "Plan"
+      ]
     }
   },
   {

package/goldens/structured-output-round.trace.json

@@ -49,7 +49,13 @@
       ],
       "output_style": "default",
       "skills": [],
-      "plugins": []
+      "plugins": [],
+      "agents": [
+        "claude",
+        "Explore",
+        "general-purpose",
+        "Plan"
+      ]
     }
   },
   {

package/goldens/subagent-permission-round.trace.json

@@ -48,7 +48,13 @@
       ],
       "output_style": "default",
       "skills": [],
-      "plugins": []
+      "plugins": [],
+      "agents": [
+        "claude",
+        "Explore",
+        "general-purpose",
+        "Plan"
+      ]
     }
   },
   {
@@ -65,7 +71,8 @@
             "name": "Agent",
             "input": {
               "description": "permission probe",
-              "prompt": "child probe text"
+              "prompt": "child probe text",
+              "run_in_background": false
             }
           }
         ]
@@ -75,6 +82,23 @@
   {
     "sequence": 2,
     "direction": "runtime-to-host",
+    "kind": "system/task_started",
+    "payload": {
+      "type": "system",
+      "subtype": "task_started",
+      "task_id": "TASKID_1",
+      "tool_use_id": "agent-call-1",
+      "description": "permission probe",
+      "subagent_type": "general-purpose",
+      "is_backgrounded": false,
+      "spawn_depth": 1,
+      "task_type": "local_agent",
+      "prompt": "child probe text"
+    }
+  },
+  {
+    "sequence": 3,
+    "direction": "runtime-to-host",
     "kind": "assistant",
     "payload": {
       "type": "assistant",
@@ -83,7 +107,7 @@
           {
             "type": "tool_use",
             "id": "child-call-1",
-            "name": "ReadNotifications",
+            "name": "ListAgents",
             "input": {}
           }
         ]
@@ -92,7 +116,25 @@
     }
   },
   {
-    "sequence": 3,
+    "sequence": 4,
+    "direction": "runtime-to-host",
+    "kind": "system/task_progress",
+    "payload": {
+      "type": "system",
+      "subtype": "task_progress",
+      "task_id": "TASKID_1",
+      "tool_use_id": "agent-call-1",
+      "description": "permission probe",
+      "subagent_type": "general-purpose",
+      "usage": {
+        "total_tokens": "<total_tokens>",
+        "tool_uses": 1
+      },
+      "last_tool_name": "ListAgents"
+    }
+  },
+  {
+    "sequence": 5,
     "direction": "runtime-to-host",
     "kind": "user",
     "payload": {
@@ -102,7 +144,7 @@
           {
             "type": "tool_result",
             "tool_use_id": "child-call-1",
-            "content": "{\"notifications\":[],\"remaining\":0}"
+            "content": "{\"listing\":\"- agent:<uuid>:<uuid> [agent/winter-agent] status=running mode=default\"}"
           }
         ]
       },
@@ -110,7 +152,7 @@
     }
   },
   {
-    "sequence": 4,
+    "sequence": 6,
     "direction": "runtime-to-host",
     "kind": "assistant",
     "payload": {
@@ -127,7 +169,39 @@
     }
   },
   {
-    "sequence": 5,
+    "sequence": 7,
+    "direction": "runtime-to-host",
+    "kind": "system/task_updated",
+    "payload": {
+      "type": "system",
+      "subtype": "task_updated",
+      "task_id": "TASKID_1",
+      "patch": {
+        "status": "completed",
+        "end_time": "<end_time>"
+      }
+    }
+  },
+  {
+    "sequence": 8,
+    "direction": "runtime-to-host",
+    "kind": "system/task_notification",
+    "payload": {
+      "type": "system",
+      "subtype": "task_notification",
+      "task_id": "TASKID_1",
+      "tool_use_id": "agent-call-1",
+      "status": "completed",
+      "output_file": "/winter-fixture-tasks/TASKID_1.output",
+      "summary": "child finished after its own tool call",
+      "usage": {
+        "total_tokens": "<total_tokens>",
+        "tool_uses": 1
+      }
+    }
+  },
+  {
+    "sequence": 9,
     "direction": "runtime-to-host",
     "kind": "user",
     "payload": {
@@ -137,14 +211,14 @@
           {
             "type": "tool_result",
             "tool_use_id": "agent-call-1",
-            "content": "{\"agentId\":\"<scrubbed>\",\"content\":[{\"type\":\"text\",\"text\":\"child finished after its own tool call\"}],\"totalToolUseCount\":1,\"totalDurationMs\":\"<scrubbed>\",\"resolvedModel\":\"winter-test/echo\",\"prompt\":\"child probe text\"}"
+            "content": "{\"agentId\":\"<scrubbed>\",\"agentType\":\"general-purpose\",\"content\":[{\"type\":\"text\",\"text\":\"child finished after its own tool call\"}],\"totalToolUseCount\":1,\"totalDurationMs\":\"<scrubbed>\",\"resolvedModel\":\"winter-test/echo\",\"prompt\":\"child probe text\"}"
           }
         ]
       }
     }
   },
   {
-    "sequence": 6,
+    "sequence": 10,
     "direction": "runtime-to-host",
     "kind": "assistant",
     "payload": {
@@ -160,7 +234,7 @@
     }
   },
   {
-    "sequence": 7,
+    "sequence": 11,
     "direction": "runtime-to-host",
     "kind": "result",
     "payload": {