codex-chatgpt-control 0.3.0-alpha.1 → 0.5.0-alpha.1

package/CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## 0.5.0-alpha.1
+
+- Adds `experience.detect/open`, `configuration.inspect/apply`, and the Work task lifecycle command group.
+- Adds scoped Chat/Work selector profiles, strict configuration postcondition verification, and sanitized profile fixtures.
+- Adds runner/Responses experience and configuration inputs plus milestone events.
+- Preserves existing `mode`, `modes.set/get`, commands, package imports, and wire fields for backward compatibility.
+
 ## 0.3.0-alpha.1
 
 - Hardens mode-menu detection and selection against thread/sidebar action menus, with locale-registry-backed thread-action vetoes and container-scoped menu enumeration.

package/README.md

@@ -1,13 +1,13 @@
 # codex-chatgpt-control
 
-TypeScript runtime for Codex agents controlling visible ChatGPT web sessions through a compatible browser bridge.
+TypeScript runtime for controlling visible ChatGPT Chat and Work through a compatible browser bridge.
 
 Unofficial project: not affiliated with, endorsed by, or sponsored by OpenAI. This is not an OpenAI API wrapper and does not call hidden or private ChatGPT endpoints. Browser-required calls need a visible session and should fail with a clear machine-readable reason when the bridge is unavailable.
 
 ## Install
 
 ```bash
-npm install codex-chatgpt-control
+npm install codex-chatgpt-control@next
 ```
 
 ## Usage
@@ -24,10 +24,50 @@ const reviewer = chatgpt.agent({
 const result = await chatgpt.runner.run(reviewer, {
   input: "Review this design.",
   thread: { type: "new" },
+  experience: "chat",
   response: { format: "markdown" }
 });
 ```
 
+Inspect the visible surface and apply verified configuration:
+
+```ts
+const surface = await chatgpt.experience.detect();
+const capabilities = await chatgpt.configuration.inspect();
+
+await chatgpt.configuration.apply({
+  experience: "work",
+  desired: {
+    model: "GPT-5.6 Sol",
+    effort: "High",
+    speed: "Standard"
+  },
+  strict: true
+});
+```
+
+Start a fresh Work task once, then poll or steer it:
+
+```ts
+await chatgpt.work.start({
+  prompt: "Produce a decision-ready implementation brief.",
+  newTask: true,
+  wait: false,
+  read: false
+});
+
+await chatgpt.work.status({ includeArtifacts: true });
+await chatgpt.work.steer({
+  prompt: "Add a prioritized migration sequence.",
+  wait: false,
+  read: false
+});
+```
+
+Legacy `mode` inputs and `modes.set/get` remain supported. New code should use
+`experience` and strict `configuration` because Chat and Work expose different
+nested axes.
+
 Reuse a user-open ChatGPT thread without replacing the tab:
 
 ```ts

package/contracts/v1/fixtures/backend-capabilities.json

@@ -26,6 +26,15 @@
     "describe",
     "help",
     "session.bootstrap",
+    "experience.detect",
+    "experience.open",
+    "configuration.inspect",
+    "configuration.apply",
+    "work.start",
+    "work.status",
+    "work.wait",
+    "work.steer",
+    "work.readLatest",
     "threads.new",
     "threads.search",
     "threads.open",
@@ -34,6 +43,7 @@
     "messages.ask",
     "messages.wait",
     "messages.readLatest",
+    "messages.status",
     "messages.waitAndRead",
     "artifacts.listLatest",
     "artifacts.wait",

package/contracts/v1/fixtures/command-descriptors.json

@@ -4,7 +4,10 @@
   "result": [
     {
       "args": {
+        "configuration": "optional surface-aware visible configuration",
         "existingTab": "true or explicit policy to claim a user-open Chrome tab instead of opening a replacement",
+        "experience": "optional chat or work surface",
+        "mode": "legacy visible mode preference",
         "prompt": "message to send or workflow-specific input",
         "report": "optional redacted report settings",
         "thread": "optional thread selector"
@@ -34,7 +37,10 @@
     },
     {
       "args": {
+        "configuration": "optional surface-aware visible configuration",
         "existingTab": "true or explicit policy to claim a user-open Chrome tab instead of opening a replacement",
+        "experience": "optional chat or work surface",
+        "mode": "legacy visible mode preference",
         "prompt": "message to send or workflow-specific input",
         "report": "optional redacted report settings",
         "thread": "optional thread selector"
@@ -65,7 +71,9 @@
     },
     {
       "args": {
+        "configuration": "optional strict visible configuration such as { intelligence: \"Pro\" } or Work model/effort/speed",
         "existingTab": "true or explicit policy to claim a user-open Chrome tab instead of opening a replacement",
+        "experience": "optional chat or work surface",
         "files": "absolute local file paths to attach before submitting",
         "mode": "optional visible mode selection, e.g. { model: \"Pro\" } or { intelligence: \"Pro\", modelVersion: \"5.4\" }",
         "prompt": "message to send after files are attached",
@@ -89,17 +97,21 @@
       },
       "defaultTimeoutMs": 120000,
       "examples": [
-        "await chatgpt.askWithFiles({ thread: { type: \"url\", url: \"https://chatgpt.com/c/<conversation-id>\" }, existingTab: true, mode: { model: \"Pro\" }, files: [\"/absolute/host/path/brief.md\"], prompt: \"Summarize this.\", wait: true, read: { format: \"markdown\" } });"
+        "await chatgpt.askWithFiles({ thread: { type: \"url\", url: \"https://chatgpt.com/c/<conversation-id>\" }, existingTab: true, experience: \"chat\", configuration: { intelligence: \"Pro\" }, files: [\"/absolute/host/path/brief.md\"], prompt: \"Summarize this.\", wait: true, read: { format: \"markdown\" } });",
+        "await chatgpt.askWithFiles({ mode: { model: \"Pro\" }, files: [\"/absolute/host/path/image.jpg\"], prompt: \"Describe this image.\", wait: true });"
       ],
       "layer": "workflow",
       "name": "askWithFiles",
       "retryPolicy": "Return structured CommandResult failures; do not resubmit prompts unless the sequence policy permits unmatched-turn recovery.",
       "risk": "medium",
-      "summary": "Attach absolute local file paths, optionally set mode, ask, wait, and read."
+      "summary": "Attach absolute local file paths, optionally open/configure Chat or Work, ask, wait, and read."
     },
     {
       "args": {
+        "configuration": "optional surface-aware visible configuration",
         "existingTab": "true or explicit policy to claim a user-open Chrome tab instead of opening a replacement",
+        "experience": "optional chat or work surface",
+        "mode": "legacy visible mode preference",
         "prompt": "message to send or workflow-specific input",
         "report": "optional redacted report settings",
         "thread": "optional thread selector"
@@ -129,7 +141,10 @@
     },
     {
       "args": {
+        "configuration": "optional surface-aware visible configuration",
         "existingTab": "true or explicit policy to claim a user-open Chrome tab instead of opening a replacement",
+        "experience": "optional chat or work surface",
+        "mode": "legacy visible mode preference",
         "prompt": "message to send or workflow-specific input",
         "report": "optional redacted report settings",
         "thread": "optional thread selector"
@@ -159,7 +174,10 @@
     },
     {
       "args": {
+        "configuration": "optional surface-aware visible configuration",
         "existingTab": "true or explicit policy to claim a user-open Chrome tab instead of opening a replacement",
+        "experience": "optional chat or work surface",
+        "mode": "legacy visible mode preference",
         "prompt": "message to send or workflow-specific input",
         "report": "optional redacted report settings",
         "thread": "optional thread selector"
@@ -179,18 +197,21 @@
       },
       "defaultTimeoutMs": 120000,
       "examples": [
-        "const agent = chatgpt.agent({ name: \"reviewer\", instructions: \"Review deeply.\" }); await chatgpt.runner.run(agent, { input: \"Review this.\", thread: { type: \"new\" } });",
+        "const agent = chatgpt.agent({ name: \"reviewer\", instructions: \"Review deeply.\" }); await chatgpt.runner.run(agent, { input: \"Review this.\", thread: { type: \"new\" }, experience: \"chat\" });",
         "await chatgpt.runner.run(agent, { input: \"Continue.\", thread: { type: \"url\", url: \"https://chatgpt.com/c/<conversation-id>\" }, existingTab: true });"
       ],
       "layer": "workflow",
       "name": "runner.run",
       "retryPolicy": "Return structured CommandResult failures; do not resubmit prompts unless the sequence policy permits unmatched-turn recovery.",
       "risk": "medium",
-      "summary": "Agents-style facade: run a visible ChatGPT browser-control agent against input, files, thread, existing-tab, mode, and response options."
+      "summary": "Agents-style facade: run a visible ChatGPT agent against input, files, thread, existing-tab, experience, configuration, legacy mode, and response options."
     },
     {
       "args": {
+        "configuration": "optional surface-aware visible configuration",
         "existingTab": "true or explicit policy to claim a user-open Chrome tab instead of opening a replacement",
+        "experience": "optional chat or work surface",
+        "mode": "legacy visible mode preference",
         "prompt": "message to send or workflow-specific input",
         "report": "optional redacted report settings",
         "thread": "optional thread selector"
@@ -220,7 +241,42 @@
     },
     {
       "args": {
+        "configuration": "optional Work model, effort, and speed values",
+        "files": "optional absolute local file paths",
+        "newTask": "start from a blank Work task; defaults to true",
+        "prompt": "visible Work task prompt",
+        "read": "optional response capture behavior",
+        "wait": "optional wait behavior"
+      },
+      "blockers": [
+        "browser_bridge_unavailable",
+        "login_required",
+        "captcha",
+        "rate_limit",
+        "selector_drift"
+      ],
+      "defaults": {
+        "read": {
+          "format": "markdown"
+        },
+        "wait": true
+      },
+      "defaultTimeoutMs": 120000,
+      "examples": [
+        "await chatgpt.work.start({ prompt: \"Analyze the attached package.\", files: [\"/absolute/path/package.tgz\"], configuration: { model: \"GPT-5.6 Sol\", effort: \"High\", speed: \"Fast\" }, wait: true, read: { format: \"markdown\" } });"
+      ],
+      "layer": "workflow",
+      "name": "work.start",
+      "retryPolicy": "Return structured CommandResult failures; do not resubmit prompts unless the sequence policy permits unmatched-turn recovery.",
+      "risk": "medium",
+      "summary": "Open Work, optionally apply model/effort/speed configuration, submit one task, and optionally wait/read without blind resubmission."
+    },
+    {
+      "args": {
+        "configuration": "optional surface-aware visible configuration",
         "existingTab": "true or explicit policy to claim a user-open Chrome tab instead of opening a replacement",
+        "experience": "optional chat or work surface",
+        "mode": "legacy visible mode preference",
         "prompt": "message to send or workflow-specific input",
         "report": "optional redacted report settings",
         "thread": "optional thread selector"
@@ -248,7 +304,10 @@
     },
     {
       "args": {
+        "configuration": "optional surface-aware visible configuration",
         "existingTab": "true or explicit policy to claim a user-open Chrome tab instead of opening a replacement",
+        "experience": "optional chat or work surface",
+        "mode": "legacy visible mode preference",
         "prompt": "message to send or workflow-specific input",
         "report": "optional redacted report settings",
         "thread": "optional thread selector"
@@ -540,6 +599,178 @@
       "risk": "low",
       "summary": "Attach to ChatGPT in Chrome and detect login/blocker state."
     },
+    {
+      "args": {
+        "timeoutMs": "optional timeout for attaching to the ChatGPT page"
+      },
+      "blockers": [
+        "browser_bridge_unavailable",
+        "login_required",
+        "selector_drift"
+      ],
+      "defaults": {},
+      "defaultTimeoutMs": 30000,
+      "examples": [
+        "await chatgpt.experience.detect();"
+      ],
+      "layer": "primitive",
+      "name": "experience.detect",
+      "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
+      "risk": "low",
+      "summary": "Detect whether the scoped visible composer is Chat, Work, or unknown and return its selector profile."
+    },
+    {
+      "args": {
+        "experience": "chat or work",
+        "timeoutMs": "optional verified-switch timeout"
+      },
+      "blockers": [
+        "browser_bridge_unavailable",
+        "login_required",
+        "selector_drift"
+      ],
+      "defaults": {},
+      "defaultTimeoutMs": 30000,
+      "examples": [
+        "await chatgpt.experience.open({ experience: \"work\" });"
+      ],
+      "layer": "primitive",
+      "name": "experience.open",
+      "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
+      "risk": "medium",
+      "summary": "Open Chat or Work and verify the target surface from scoped composer evidence."
+    },
+    {
+      "args": {
+        "experience": "optional expected chat or work surface",
+        "includeOptions": "open visible axis menus to enumerate options; defaults to true",
+        "timeoutMs": "optional inspection timeout"
+      },
+      "blockers": [
+        "browser_bridge_unavailable",
+        "login_required",
+        "selector_drift"
+      ],
+      "defaults": {},
+      "defaultTimeoutMs": 30000,
+      "examples": [
+        "await chatgpt.configuration.inspect({ experience: \"work\" });"
+      ],
+      "layer": "primitive",
+      "name": "configuration.inspect",
+      "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
+      "risk": "low",
+      "summary": "Inspect the active Chat or Work selector profile, axes, options, and selected values."
+    },
+    {
+      "args": {
+        "desired": "model, intelligence, effort, speed, and/or modelVersion values",
+        "experience": "optional target chat or work surface",
+        "strict": "fail when visible postconditions are not verified; defaults to true",
+        "timeoutMs": "optional selection timeout"
+      },
+      "blockers": [
+        "browser_bridge_unavailable",
+        "login_required",
+        "selector_drift"
+      ],
+      "defaults": {},
+      "defaultTimeoutMs": 30000,
+      "examples": [
+        "await chatgpt.configuration.apply({ experience: \"work\", desired: { model: \"GPT-5.6 Sol\", effort: \"High\", speed: \"Fast\" }, strict: true });"
+      ],
+      "layer": "primitive",
+      "name": "configuration.apply",
+      "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
+      "risk": "medium",
+      "summary": "Apply model/intelligence/effort/speed configuration one axis at a time and verify the final visible state."
+    },
+    {
+      "args": {
+        "includeArtifacts": "include the latest visible artifact inventory; defaults to false for lightweight polling",
+        "maxPreviewChars": "maximum assistant preview characters"
+      },
+      "blockers": [
+        "browser_bridge_unavailable",
+        "login_required",
+        "selector_drift"
+      ],
+      "defaults": {},
+      "defaultTimeoutMs": 5000,
+      "examples": [
+        "await chatgpt.work.status();"
+      ],
+      "layer": "primitive",
+      "name": "work.status",
+      "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
+      "risk": "medium",
+      "summary": "Read current Work task progress and optionally enumerate visible artifacts."
+    },
+    {
+      "args": {
+        "pollMs": "polling interval",
+        "stableMs": "required response stability window",
+        "timeoutMs": "optional Work wait timeout"
+      },
+      "blockers": [
+        "browser_bridge_unavailable",
+        "login_required",
+        "selector_drift"
+      ],
+      "defaults": {},
+      "defaultTimeoutMs": 120000,
+      "examples": [
+        "await chatgpt.work.wait({ timeoutMs: 600000 });"
+      ],
+      "layer": "primitive",
+      "name": "work.wait",
+      "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
+      "risk": "low",
+      "summary": "Wait for the current Work task response to stabilize."
+    },
+    {
+      "args": {
+        "prompt": "visible steering message for the current task",
+        "read": "optional response capture behavior",
+        "wait": "optional wait behavior"
+      },
+      "blockers": [
+        "browser_bridge_unavailable",
+        "login_required",
+        "selector_drift"
+      ],
+      "defaults": {},
+      "defaultTimeoutMs": 120000,
+      "examples": [
+        "await chatgpt.work.steer({ prompt: \"Focus the comparison on deployment ergonomics.\" });"
+      ],
+      "layer": "primitive",
+      "name": "work.steer",
+      "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
+      "risk": "medium",
+      "summary": "Submit a visible steering message to the current Work task without opening a new task."
+    },
+    {
+      "args": {
+        "format": "markdown, normalized_text, visible_text, html, blocks, or all",
+        "maxChars": "optional capture limit"
+      },
+      "blockers": [
+        "browser_bridge_unavailable",
+        "login_required",
+        "selector_drift"
+      ],
+      "defaults": {},
+      "defaultTimeoutMs": 30000,
+      "examples": [
+        "await chatgpt.work.readLatest({ format: \"markdown\" });"
+      ],
+      "layer": "primitive",
+      "name": "work.readLatest",
+      "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
+      "risk": "medium",
+      "summary": "Read the latest response from the current Work task."
+    },
     {
       "args": {},
       "blockers": [
@@ -695,6 +926,26 @@
       "risk": "medium",
       "summary": "Read the latest message as Markdown, normalized text, blocks, or HTML."
     },
+    {
+      "args": {
+        "maxPreviewChars": "maximum latest-assistant preview characters to return; defaults to 240"
+      },
+      "blockers": [
+        "browser_bridge_unavailable",
+        "login_required",
+        "captcha",
+        "rate_limit",
+        "selector_drift"
+      ],
+      "defaults": {},
+      "defaultTimeoutMs": 5000,
+      "examples": [],
+      "layer": "primitive",
+      "name": "messages.status",
+      "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
+      "risk": "medium",
+      "summary": "Read a compact latest-assistant progress snapshot without treating partial text as complete."
+    },
     {
       "args": {},
       "blockers": [
@@ -967,7 +1218,7 @@
       "name": "modes.set",
       "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
       "risk": "medium",
-      "summary": "Select a visible model, intelligence, effort, or nested model-version candidate when unambiguous."
+      "summary": "Legacy compatibility command: select a visible model/intelligence/effort/version candidate with historical warning-oriented verification."
     },
     {
       "args": {
@@ -988,7 +1239,7 @@
       "name": "modes.get",
       "retryPolicy": "Return structured CommandResult failures; retry only when the blocker is recoverable and no duplicate prompt will be submitted.",
       "risk": "low",
-      "summary": "Read the mode labels shown on the visible composer controls without changing them."
+      "summary": "Legacy compatibility command: read mode labels shown on visible composer controls without changing them."
     },
     {
       "args": {},

package/contracts/v1/fixtures/describe-runner-run.json

@@ -1,6 +1,9 @@
 {
   "args": {
+    "configuration": "optional surface-aware visible configuration",
     "existingTab": "true or explicit policy to claim a user-open Chrome tab instead of opening a replacement",
+    "experience": "optional chat or work surface",
+    "mode": "legacy visible mode preference",
     "prompt": "message to send or workflow-specific input",
     "report": "optional redacted report settings",
     "thread": "optional thread selector"
@@ -20,12 +23,12 @@
   },
   "defaultTimeoutMs": 120000,
   "examples": [
-    "const agent = chatgpt.agent({ name: \"reviewer\", instructions: \"Review deeply.\" }); await chatgpt.runner.run(agent, { input: \"Review this.\", thread: { type: \"new\" } });",
+    "const agent = chatgpt.agent({ name: \"reviewer\", instructions: \"Review deeply.\" }); await chatgpt.runner.run(agent, { input: \"Review this.\", thread: { type: \"new\" }, experience: \"chat\" });",
     "await chatgpt.runner.run(agent, { input: \"Continue.\", thread: { type: \"url\", url: \"https://chatgpt.com/c/<conversation-id>\" }, existingTab: true });"
   ],
   "layer": "workflow",
   "name": "runner.run",
   "retryPolicy": "Return structured CommandResult failures; do not resubmit prompts unless the sequence policy permits unmatched-turn recovery.",
   "risk": "medium",
-  "summary": "Agents-style facade: run a visible ChatGPT browser-control agent against input, files, thread, existing-tab, mode, and response options."
+  "summary": "Agents-style facade: run a visible ChatGPT agent against input, files, thread, existing-tab, experience, configuration, legacy mode, and response options."
 }

package/contracts/v1/fixtures/doctor-scenario-preflight.json

@@ -71,7 +71,7 @@
         "localization": {
           "details": {
             "englishCanonicalPresent": true,
-            "labelCandidateCount": 1542,
+            "labelCandidateCount": 1544,
             "missingModeOptionIds": [],
             "missingToolIds": [],
             "modeOptionIds": [
@@ -85,15 +85,47 @@
               "pro"
             ],
             "pageAvailable": false,
+            "proAliases": [
+              "Pro",
+              "Pro Extended",
+              "Pro • Extended",
+              "Extended Pro",
+              "專業",
+              "احترافي",
+              "Про",
+              "حرفه‌ای",
+              "Պրո",
+              "ಪ್ರೊ",
+              "Profesionalus",
+              "专业",
+              "പ്രോ",
+              "ਪ੍ਰੋ",
+              "प्रो",
+              "ప్రో",
+              "প্রো",
+              "ப்ரோ"
+            ],
             "requiredKeysMissing": [],
+            "runningStateLabelCoverage": {
+              "nonEnglishLocaleCount": 64,
+              "nonEnglishStopControlLocaleCount": 64,
+              "nonEnglishStoppedAssistantLocaleCount": 0,
+              "registeredLocaleCount": 65,
+              "stopControlCandidateCount": 65,
+              "stopControlLocaleCount": 65,
+              "stoppedAssistantCandidateCount": 3,
+              "stoppedAssistantLocaleCount": 1,
+              "support": "partial"
+            },
             "runtimeSelectorCoverage": "registry_only_stage_2",
+            "supportsProExtendedAlias": true,
             "toolIds": [
               "web_search",
               "deep_research",
               "create_image"
             ]
           },
-          "message": "The locale registry is loaded; localized runtime selector coverage is registry-only in Stage 2 and not fully proven.",
+          "message": "The locale registry is loaded; localized runtime selector coverage is registry-only in Stage 2 and running-state labels are partially localized.",
           "status": "unknown"
         },
         "reports": {

package/contracts/v1/fixtures/help-root.json

@@ -1,6 +1,6 @@
 {
   "ok": true,
   "requestId": "req_help",
-  "result": "ChatGPT browser-control SDK commands\n\nworkflow:\n- ask: Ask ChatGPT in a new or selected thread, optionally with files, wait/read, downloads, and reports.\n- askInThread: Open or claim an existing thread by URL, conversation id, title, or search query, then ask and read.\n- askWithFiles: Attach absolute local file paths, optionally set mode, ask, wait, and read.\n- askAndDownload: Ask ChatGPT to produce a visible downloadable output and save the latest exposed file.\n- runMessages: Run sequential prompts where later prompts can use earlier step data.\n- runner.run: Agents-style facade: run a visible ChatGPT browser-control agent against input, files, thread, existing-tab, mode, and response options.\n- responses.create: Narrow Responses-shaped adapter over the visible ChatGPT browser-control runner; rejects unsupported API-only fields before prompt submission.\n- copyLatest: Copy or DOM-read the latest assistant response with Markdown-first fidelity.\n- runPlan: Execute an inline SequencePlan or named macro through the existing sequence engine.\n- new-ask-read: Named macro: open a new thread, ask, wait, and read Markdown.\n- find-open-ask-read: Named macro: search history, open the first match, ask, wait, and read Markdown.\n- find-open-copy-latest: Named macro: search history, open the first match, and copy/read the latest response.\n- attach-ask-read: Named macro: open a new thread, attach files, ask, wait, and read Markdown.\n- ask-and-download: Named macro: ask in a new thread and download the latest file affordance.\n- two-turn: Named macro: run two sequential prompts in a new thread.\ndiagnostic:\n- doctor-upload: Named macro: preflight bridge, login, and upload permission remediation.\n- doctor: Preflight browser bridge, login, upload, local files, existing-tab, artifact, localization, report, and selector readiness.\nreport:\n- redacted-run-report: Named macro: create a redacted report for a supplied CommandResult.\n- createReport: Write a durable redacted run report for a command result.\nprimitive:\n- session.bootstrap: Attach to ChatGPT in Chrome and detect login/blocker state.\n- threads.new: Open a new ChatGPT thread.\n- threads.search: Search visible ChatGPT history by query.\n- threads.open: Open a thread by URL, conversation id, title, or search result.\n- messages.compose: Fill the composer without submitting.\n- messages.submit: Submit the current composer contents.\n- messages.ask: Compose, submit, optionally wait, and optionally read.\n- messages.wait: Wait for the latest assistant response to stabilize.\n- messages.readLatest: Read the latest message as Markdown, normalized text, blocks, or HTML.\n- messages.waitAndRead: Wait for completion and read the latest message.\n- artifacts.listLatest: Detect the latest visible generated ChatGPT artifact, such as an image-only result.\n- artifacts.wait: Wait for a visible generated ChatGPT artifact to appear and stabilize.\n- artifacts.downloadLatest: Download or save the latest visible generated ChatGPT artifact.\n- files.preflight: Validate local file paths, size limits, duplicates, zero-byte files, and extension-based MIME/category guesses without opening ChatGPT.\n- files.attach: Attach absolute local file paths through visible ChatGPT upload controls.\n- files.downloadLatest: Download the latest visible ChatGPT file affordance.\n- projects.sources.list: Open or claim a visible ChatGPT Project Sources tab and list source names/statuses without source contents.\n- projects.sources.planAdd: Dry-run an append-only Project Sources file add from explicit local files without opening ChatGPT.\n- projects.sources.add: Append explicit local files to a visible ChatGPT Project Sources list after confirmMutation: true.\n- response.copy: Click Copy response and return clipboard Markdown, with DOM fallback.\n- modes.set: Select a visible model, intelligence, effort, or nested model-version candidate when unambiguous.\n- modes.get: Read the mode labels shown on the visible composer controls without changing them.\n- tools.select: Select a visible ChatGPT tool when unambiguous.",
+  "result": "ChatGPT browser-control SDK commands\n\nworkflow:\n- ask: Ask ChatGPT in a new or selected thread, optionally with files, wait/read, downloads, and reports.\n- askInThread: Open or claim an existing thread by URL, conversation id, title, or search query, then ask and read.\n- askWithFiles: Attach absolute local file paths, optionally open/configure Chat or Work, ask, wait, and read.\n- askAndDownload: Ask ChatGPT to produce a visible downloadable output and save the latest exposed file.\n- runMessages: Run sequential prompts where later prompts can use earlier step data.\n- runner.run: Agents-style facade: run a visible ChatGPT agent against input, files, thread, existing-tab, experience, configuration, legacy mode, and response options.\n- responses.create: Narrow Responses-shaped adapter over the visible ChatGPT browser-control runner; rejects unsupported API-only fields before prompt submission.\n- work.start: Open Work, optionally apply model/effort/speed configuration, submit one task, and optionally wait/read without blind resubmission.\n- copyLatest: Copy or DOM-read the latest assistant response with Markdown-first fidelity.\n- runPlan: Execute an inline SequencePlan or named macro through the existing sequence engine.\n- new-ask-read: Named macro: open a new thread, ask, wait, and read Markdown.\n- find-open-ask-read: Named macro: search history, open the first match, ask, wait, and read Markdown.\n- find-open-copy-latest: Named macro: search history, open the first match, and copy/read the latest response.\n- attach-ask-read: Named macro: open a new thread, attach files, ask, wait, and read Markdown.\n- ask-and-download: Named macro: ask in a new thread and download the latest file affordance.\n- two-turn: Named macro: run two sequential prompts in a new thread.\ndiagnostic:\n- doctor-upload: Named macro: preflight bridge, login, and upload permission remediation.\n- doctor: Preflight browser bridge, login, upload, local files, existing-tab, artifact, localization, report, and selector readiness.\nreport:\n- redacted-run-report: Named macro: create a redacted report for a supplied CommandResult.\n- createReport: Write a durable redacted run report for a command result.\nprimitive:\n- session.bootstrap: Attach to ChatGPT in Chrome and detect login/blocker state.\n- experience.detect: Detect whether the scoped visible composer is Chat, Work, or unknown and return its selector profile.\n- experience.open: Open Chat or Work and verify the target surface from scoped composer evidence.\n- configuration.inspect: Inspect the active Chat or Work selector profile, axes, options, and selected values.\n- configuration.apply: Apply model/intelligence/effort/speed configuration one axis at a time and verify the final visible state.\n- work.status: Read current Work task progress and optionally enumerate visible artifacts.\n- work.wait: Wait for the current Work task response to stabilize.\n- work.steer: Submit a visible steering message to the current Work task without opening a new task.\n- work.readLatest: Read the latest response from the current Work task.\n- threads.new: Open a new ChatGPT thread.\n- threads.search: Search visible ChatGPT history by query.\n- threads.open: Open a thread by URL, conversation id, title, or search result.\n- messages.compose: Fill the composer without submitting.\n- messages.submit: Submit the current composer contents.\n- messages.ask: Compose, submit, optionally wait, and optionally read.\n- messages.wait: Wait for the latest assistant response to stabilize.\n- messages.readLatest: Read the latest message as Markdown, normalized text, blocks, or HTML.\n- messages.status: Read a compact latest-assistant progress snapshot without treating partial text as complete.\n- messages.waitAndRead: Wait for completion and read the latest message.\n- artifacts.listLatest: Detect the latest visible generated ChatGPT artifact, such as an image-only result.\n- artifacts.wait: Wait for a visible generated ChatGPT artifact to appear and stabilize.\n- artifacts.downloadLatest: Download or save the latest visible generated ChatGPT artifact.\n- files.preflight: Validate local file paths, size limits, duplicates, zero-byte files, and extension-based MIME/category guesses without opening ChatGPT.\n- files.attach: Attach absolute local file paths through visible ChatGPT upload controls.\n- files.downloadLatest: Download the latest visible ChatGPT file affordance.\n- projects.sources.list: Open or claim a visible ChatGPT Project Sources tab and list source names/statuses without source contents.\n- projects.sources.planAdd: Dry-run an append-only Project Sources file add from explicit local files without opening ChatGPT.\n- projects.sources.add: Append explicit local files to a visible ChatGPT Project Sources list after confirmMutation: true.\n- response.copy: Click Copy response and return clipboard Markdown, with DOM fallback.\n- modes.set: Legacy compatibility command: select a visible model/intelligence/effort/version candidate with historical warning-oriented verification.\n- modes.get: Legacy compatibility command: read mode labels shown on visible composer controls without changing them.\n- tools.select: Select a visible ChatGPT tool when unambiguous.",
   "schemaVersion": "chatgpt.browser_control.backend_response.v1"
 }

package/contracts/v1/fixtures/output-json-parse-success.json

@@ -9,11 +9,14 @@
       "url": "https://chatgpt.com/"
     },
     "data": {
+      "completionState": "complete",
       "finalOutput": {
         "score": 1,
         "verdict": "ok"
       },
+      "generationActive": false,
       "outputText": "{\"verdict\":\"ok\",\"score\":1}",
+      "submissionState": "submitted",
       "thread": {
         "title": "ChatGPT",
         "url": "https://chatgpt.com/"
@@ -67,8 +70,10 @@
     ],
     "output_text": "{\"verdict\":\"ok\",\"score\":1}",
     "state": {
+      "completionState": "complete",
       "id": "run_fixed",
       "resumable": false,
+      "submissionState": "submitted",
       "thread": {
         "title": "ChatGPT",
         "url": "https://chatgpt.com/"
@@ -108,8 +113,12 @@
         "command": "messages.ask",
         "dataPreview": {
           "complete": true,
+          "completionState": "complete",
+          "generationActive": false,
+          "generationSignals": [],
           "prompt": "Return a JSON verdict.",
           "responseText": "{\"verdict\":\"ok\",\"score\":1}",
+          "submissionState": "submitted",
           "title": "ChatGPT"
         },
         "endedAt": "2026-06-06T00:00:00.000Z",