oh-my-opencode 5.0.0-beta.85 → 5.0.0-beta.87

package/dist/index.js

@@ -20057,26 +20057,20 @@ var AGENT_MODEL_REQUIREMENTS = {
     fallbackChain: [
       { providers: ["kimi-for-coding"], model: "kimi-for-coding-highspeed", variant: "off" },
       { providers: ["openai", "chatgpt-subscription"], model: "gpt-6-luna-fast", variant: "low" },
-      { providers: ["deepseek"], model: "deepseek-v4-flash", variant: "max" },
+      { providers: ["deepseek"], model: "deepseek-flash", variant: "max" },
       { providers: ["opencode-go", "bailian-coding-plan"], model: "qwen3.7-plus" },
-      { providers: ["opencode-go"], model: "minimax-m3" },
-      { providers: ["minimax-coding-plan", "minimax-cn-coding-plan"], model: "MiniMax-M3" },
       { providers: ["opencode-go"], model: "minimax-m2.7" },
-      { providers: ["anthropic", "github-copilot"], model: "claude-haiku-4-5" },
-      { providers: ["openai", "chatgpt-subscription"], model: "gpt-5.4-nano" }
+      { providers: ["anthropic", "github-copilot"], model: "claude-haiku-4-5" }
     ]
   },
   explore: {
     fallbackChain: [
       { providers: ["kimi-for-coding"], model: "kimi-for-coding-highspeed", variant: "off" },
       { providers: ["openai", "chatgpt-subscription"], model: "gpt-6-luna-fast", variant: "low" },
-      { providers: ["deepseek"], model: "deepseek-v4-flash", variant: "max" },
+      { providers: ["deepseek"], model: "deepseek-flash", variant: "max" },
       { providers: ["opencode-go", "bailian-coding-plan"], model: "qwen3.7-plus" },
-      { providers: ["opencode-go"], model: "minimax-m3" },
-      { providers: ["minimax-coding-plan", "minimax-cn-coding-plan"], model: "MiniMax-M3" },
       { providers: ["opencode-go"], model: "minimax-m2.7" },
-      { providers: ["anthropic", "github-copilot"], model: "claude-haiku-4-5" },
-      { providers: ["openai", "chatgpt-subscription"], model: "gpt-5.4-nano" }
+      { providers: ["anthropic", "github-copilot"], model: "claude-haiku-4-5" }
     ]
   },
   "multimodal-looker": {
@@ -20206,15 +20200,11 @@ var CATEGORY_MODEL_REQUIREMENTS = {
   },
   "deep-low": {
     fallbackChain: [
+      { providers: ["openai", "chatgpt-subscription"], model: "gpt-6-sol-fast", variant: "medium" },
       {
         providers: ["openai", "chatgpt-subscription", "github-copilot", "opencode"],
         model: "gpt-6-sol",
         variant: "medium"
-      },
-      {
-        providers: ["openai", "chatgpt-subscription", "github-copilot", "opencode"],
-        model: "gpt-5.6-sol",
-        variant: "medium"
       }
     ]
   },
@@ -20223,7 +20213,7 @@ var CATEGORY_MODEL_REQUIREMENTS = {
       {
         providers: ["openai", "chatgpt-subscription", "github-copilot", "opencode"],
         model: "gpt-6-astra",
-        variant: "high"
+        variant: "xhigh"
       }
     ]
   },
@@ -20248,8 +20238,8 @@ var CATEGORY_MODEL_REQUIREMENTS = {
   },
   quick: {
     fallbackChain: [
-      { providers: ["chatgpt-subscription"], model: "gpt-6-luna-fast", variant: "low" },
-      { providers: ["deepseek"], model: "deepseek-v4-flash", variant: "off" },
+      { providers: ["openai", "chatgpt-subscription"], model: "gpt-6-luna-fast", variant: "low" },
+      { providers: ["deepseek"], model: "deepseek-flash", variant: "off" },
       {
         providers: ["qwen-token-plan", "alibaba-token-plan", "bailian-coding-plan"],
         model: "qwen3.6-flash",
@@ -20293,7 +20283,7 @@ var CATEGORY_MODEL_REQUIREMENTS = {
       {
         providers: ["anthropic", "anthropic-api", "github-copilot", "opencode"],
         model: "claude-opus-5-5",
-        variant: "max"
+        variant: "medium"
       },
       { providers: ["zai-coding-plan", "opencode-go"], model: "glm-5.3", variant: "max" },
       {
@@ -20304,6 +20294,7 @@ var CATEGORY_MODEL_REQUIREMENTS = {
     ]
   },
   writing: {
+    requiresAnyModel: true,
     fallbackChain: [
       {
         providers: ["anthropic", "anthropic-api", "github-copilot", "opencode"],
@@ -21915,6 +21906,22 @@ var SUPPLEMENTAL_MODEL_CAPABILITIES = {
       output: 128000
     }
   },
+  "gpt-6-sol-fast": {
+    id: "gpt-6-sol-fast",
+    family: "gpt",
+    reasoning: true,
+    temperature: false,
+    toolCall: true,
+    modalities: {
+      input: ["text", "image"],
+      output: ["text"]
+    },
+    limit: {
+      context: 1050000,
+      input: 922000,
+      output: 128000
+    }
+  },
   "gpt-6-luna": {
     id: "gpt-6-luna",
     family: "gpt-nano",
@@ -21995,6 +22002,21 @@ var SUPPLEMENTAL_MODEL_CAPABILITIES = {
       output: 128000
     }
   },
+  "deepseek-flash": {
+    id: "deepseek-flash",
+    family: "deepseek-flash",
+    reasoning: true,
+    temperature: true,
+    toolCall: true,
+    modalities: {
+      input: ["text", "image"],
+      output: ["text"]
+    },
+    limit: {
+      context: 1e6,
+      output: 384000
+    }
+  },
   "gpt-6-luna-fast": {
     id: "gpt-6-luna-fast",
     family: "gpt-nano",
@@ -23540,6 +23562,8 @@ function isRecord9(value) {
 }
 var OmoModelProfileInputSchema = z9.object({
   display_name: z9.string().optional(),
+  family: z9.enum(["daily", "geeky"]).optional(),
+  tier: z9.enum(["normal", "heavy"]).optional(),
   models: z9.array(z9.union([z9.string(), OmoFallbackModelObjectSchema])).optional()
 }).strict();
 var OmoModelProfileSchema = z9.preprocess((value) => isRecord9(value) ? normalizeLegacyModelFields(value) : value, OmoModelProfileInputSchema);
@@ -27292,7 +27316,7 @@ You are working on tasks that don't fit specific categories but require substant
 var UNSPECIFIED_HIGH_CATEGORY_CALLER_GUIDANCE = `<Selection_Gate>Use only when no specialist category fits and substantial effort spans systems/modules with broad impact. Use unspecified-low for contained moderate work.</Selection_Gate>`;
 var DEEP_LOW_CATEGORY_CALLER_GUIDANCE = `<Selection_Gate>Route here when one subsystem plus its callers holds the mechanism and the evidence, once read, leaves one right answer. Wide but mechanical work belongs here or in a quick batch. When unsure, choose deep-low: a misrouted child returns \`ESCALATE: deep-high\` after one cheap attempt; re-spawn the same brief as deep-high with its findings.</Selection_Gate>`;
 var DEEP_HIGH_CATEGORY_CALLER_GUIDANCE = `<Selection_Gate>Route here only when you can name the decision evidence cannot settle: a trade-off with no single right answer, a contract change crossing a package or process boundary, a mechanism with no in-repo pattern to copy, or correctness argued from invariants rather than observed in a test. Wide scope with easy decisions is deep-low or unspecified-high; reasoning as the deliverable is ultrabrain.</Selection_Gate>`;
-var DEEP_LOW_GATE_MODELS = ["gpt-6-sol", "gpt-5.6-sol"];
+var DEEP_LOW_GATE_MODELS = ["gpt-6-sol-fast", "gpt-6-sol"];
 var DEEP_HIGH_GATE_MODEL = "gpt-6-astra";
 var OPENAI_CATEGORIES = [
   {
@@ -27304,7 +27328,7 @@ var OPENAI_CATEGORIES = [
   },
   {
     name: "deep-low",
-    config: { model: "openai/gpt-6-sol", variant: "medium" },
+    config: { model: "openai/gpt-6-sol-fast", variant: "medium" },
     description: "Default deep lane: one goal, one deliverable, decisions the child can settle from what it reads. **3D graphics, computer/browser use, CAPTCHA, multimodal, backend, logic, and algorithm work is routed here.** Multiple goals fan out as parallel calls.",
     callerGuidance: DEEP_LOW_CATEGORY_CALLER_GUIDANCE,
     promptAppend: DEEP_LOW_CATEGORY_PROMPT_APPEND,
@@ -27313,7 +27337,7 @@ var OPENAI_CATEGORIES = [
   },
   {
     name: "deep-high",
-    config: { model: "openai/gpt-6-astra", variant: "high" },
+    config: { model: "openai/gpt-6-astra", variant: "xhigh" },
     description: "Escalation deep lane: a goal whose central decision cannot be settled from evidence alone. Same one-goal, one-deliverable contract as deep-low.",
     callerGuidance: DEEP_HIGH_CATEGORY_CALLER_GUIDANCE,
     promptAppend: DEEP_HIGH_CATEGORY_PROMPT_APPEND,
@@ -27336,7 +27360,7 @@ var OPENAI_CATEGORIES = [
   },
   {
     name: "unspecified-high",
-    config: { model: "anthropic/claude-opus-5-5", variant: "max" },
+    config: { model: "anthropic/claude-opus-5-5", variant: "medium" },
     description: "Tasks that don't fit other categories, high effort required",
     callerGuidance: UNSPECIFIED_HIGH_CATEGORY_CALLER_GUIDANCE,
     promptAppend: UNSPECIFIED_HIGH_CATEGORY_PROMPT_APPEND,
@@ -97691,7 +97715,7 @@ function getCachedVersion(options = {}) {
 // package.json
 var package_default = {
   name: "oh-my-opencode",
-  version: "5.0.0-beta.85",
+  version: "5.0.0-beta.87",
   description: "The Best AI Agent Harness - Batteries-Included OpenCode Plugin with Multi-Model Orchestration, Parallel Background Agents, and Crafted LSP/AST Tools",
   main: "./dist/index.js",
   types: "dist/index.d.ts",
@@ -97874,7 +97898,7 @@ var package_default = {
     zod: "^4.6.5"
   },
   devDependencies: {
-    "@code-yeongyu/senpi": "2026.9.23",
+    "@code-yeongyu/senpi": "2026.9.23-4",
     "@oh-my-opencode/agents-md-core": "workspace:*",
     "@oh-my-opencode/ast-grep-mcp": "workspace:*",
     "@oh-my-opencode/boulder-state": "workspace:*",
@@ -97910,22 +97934,23 @@ var package_default = {
     "@xterm/xterm": "^6.0.0",
     "bun-types": "1.4.2",
     "node-pty": "^1.1.0",
+    omowright: "github:code-yeongyu/omowright#25781c8ac04f9aae358b06551520b792ac45c6ed",
     "puppeteer-core": "^25.11.0",
     typescript: "^7.0.2"
   },
   optionalDependencies: {
-    "oh-my-opencode-darwin-arm64": "5.0.0-beta.85",
-    "oh-my-opencode-darwin-x64": "5.0.0-beta.85",
-    "oh-my-opencode-darwin-x64-baseline": "5.0.0-beta.85",
-    "oh-my-opencode-linux-arm64": "5.0.0-beta.85",
-    "oh-my-opencode-linux-arm64-musl": "5.0.0-beta.85",
-    "oh-my-opencode-linux-x64": "5.0.0-beta.85",
-    "oh-my-opencode-linux-x64-baseline": "5.0.0-beta.85",
-    "oh-my-opencode-linux-x64-musl": "5.0.0-beta.85",
-    "oh-my-opencode-linux-x64-musl-baseline": "5.0.0-beta.85",
-    "oh-my-opencode-windows-arm64": "5.0.0-beta.85",
-    "oh-my-opencode-windows-x64": "5.0.0-beta.85",
-    "oh-my-opencode-windows-x64-baseline": "5.0.0-beta.85"
+    "oh-my-opencode-darwin-arm64": "5.0.0-beta.87",
+    "oh-my-opencode-darwin-x64": "5.0.0-beta.87",
+    "oh-my-opencode-darwin-x64-baseline": "5.0.0-beta.87",
+    "oh-my-opencode-linux-arm64": "5.0.0-beta.87",
+    "oh-my-opencode-linux-arm64-musl": "5.0.0-beta.87",
+    "oh-my-opencode-linux-x64": "5.0.0-beta.87",
+    "oh-my-opencode-linux-x64-baseline": "5.0.0-beta.87",
+    "oh-my-opencode-linux-x64-musl": "5.0.0-beta.87",
+    "oh-my-opencode-linux-x64-musl-baseline": "5.0.0-beta.87",
+    "oh-my-opencode-windows-arm64": "5.0.0-beta.87",
+    "oh-my-opencode-windows-x64": "5.0.0-beta.87",
+    "oh-my-opencode-windows-x64-baseline": "5.0.0-beta.87"
   },
   overrides: {
     "@earendil-works/pi-agent-core": "0.84.2",
@@ -103766,9 +103791,9 @@ task(category="quick", load_skills=["git-master"], run_in_background=true)
 
 ## EXECUTION RULES
 - **TODO format**: \`path: <action> for <scenario-id> \u2014 verify by <check>\` encoding WHERE / WHY (which scenario it advances) / HOW / VERIFY. Exactly ONE in_progress at a time. Mark completed IMMEDIATELY \u2014 never batch.
-  - GOOD pair (test-first, ordered): \`module.test: Write FAILING case invalid-email\u2192ValidationError for S2 - verify by RED with assertion msg\` \u2192 \`src/module: Implement validateEmail() for S2 - verify by module.test GREEN + curl 400 body\`
-  - BAD: "Implement feature" / "Fix bug" / "Add tests later" / production code before its failing test \u2192 rewrite.
-- **PARALLEL**: Fire independent agent calls simultaneously via task(run_in_background=true) \u2014 NEVER wait sequentially. But NEVER parallelise RED and GREEN of the same scenario.
+  - GOOD pair (ordered): \`test/module.test: read the validateEmail cases for S2 - verify by noting intent / coverage / pass in the notepad\` \u2192 \`src/module: Implement validateEmail() for S2 - verify by curl 400 body + module.test green\`
+  - BAD: "Implement feature" / "Fix bug" / "Add tests later" \u2192 rewrite.
+- **PARALLEL**: Fire independent agent calls simultaneously via task(run_in_background=true) \u2014 NEVER wait sequentially. But READ before CHANGE, never in parallel with it.
 - **BACKGROUND FIRST**: Use task for exploration/research agents (10+ concurrent if needed).
 - **VERIFY**: Re-read request after completion. Check every scenario PASS with both artifacts captured.
 - **DELEGATE**: Don't do everything yourself \u2014 orchestrate specialized agents for their strengths.
@@ -103799,10 +103824,10 @@ BEFORE writing ANY code, define realistic scenarios sized to the change \u2014 *
 
 Each scenario MUST specify, upfront:
 - Pass condition as a binary observable ("returns 200 + body matches schema"), not "should work".
-- The REAL surface that proves it: tmux transcript, curl status+body, browser/Playwright assertion, computer-use action log, CLI stdout, parsed config dump, DB state diff. Asserting "tests pass" alone is NOT evidence.
-- The cheapest faithful proof: a test file + test id at a code seam (written test-first \u2014 see TDD below), or the real-surface scenario itself when no seam exists. Prose, docs, prompt, and visual-only changes take review + real-surface QA \u2014 a test pinning their text is pretend-coverage, not proof.
+- The REAL surface that proves it: tmux transcript, curl status+body, browser (omowright) assertion, computer-use action log, CLI stdout, parsed config dump, DB state diff. Asserting "tests pass" alone is NOT evidence.
+- The existing tests that cover it (read first \u2014 see Test Decision below) and the real-surface scenario that proves it. Prose, docs, prompt, and visual-only changes take review + real-surface QA \u2014 a test pinning their text is pretend-coverage, not proof.
 
-**These scenarios are the CONTRACT.** Record them in your TODO/notepad. You are not done until every one PASSES with both pieces of evidence captured (RED\u2192GREEN proof + real-surface artifact).
+**These scenarios are the CONTRACT.** Record them in your TODO/notepad. You are not done until every one PASSES with its real-surface artifact captured and the tests the repository keeps for it green.
 
 ### Durable Notepad (survives context loss)
 
@@ -103828,12 +103853,12 @@ Every scenario requires TWO captured artifacts \u2014 both mandatory:
 
 | Artifact | Source | Captures |
 |----------|--------|----------|
-| **RED\u2192GREEN proof** | Test runner output before AND after the change | Test id + assertion message in both states |
-| **Real-surface artifact** | tmux / curl / browser / Playwright / computer-use / CLI / DB | What the user actually sees |
+| **Tests of record** | The existing suite for the area, read before the change and green after it | Intent / coverage / pass noted; stale expectations updated |
+| **Real-surface artifact** | tmux / curl / browser (omowright) / computer-use / CLI / DB | What the user actually sees |
 
 Supporting (necessary, not sufficient): build exit 0, full suite green, lsp_diagnostics clean on changed files, regression scenarios still PASS.
 
-The real-surface artifact is always required. A test is required only where a code seam exists. "tests pass" alone is NOT done, and a test pinning prose or visual text is NOT evidence.
+The real-surface artifact is always required. A new test only where the repository keeps tests for this behavior and a regression would otherwise pass unnoticed. "tests pass" alone is NOT done, and a test pinning prose or visual text is NOT evidence.
 
 <MANUAL_QA_MANDATE>
 ### YOU MUST EXECUTE MANUAL QA YOURSELF. THIS IS NOT OPTIONAL.
@@ -103847,7 +103872,7 @@ The real-surface artifact is always required. A test is required only where a co
 | Adds/modifies a CLI command | Run the command with Bash. Show the output. |
 | Changes build output | Run the build. Verify the output files exist and are correct. |
 | Modifies API behavior | Call the endpoint. Show the response. |
-| Changes UI rendering | Drive the REAL page from js eval: (1) \`new Bun.WebView()\` on Bun >= 1.4 (macOS default; Linux/Windows need installed Chrome/Chromium/Edge). (2) Otherwise, or for Chrome semantics, stealth, trace, or auth, WRITE a \`playwright-core\` script and run it from the kernel against local Chrome (\`chromium.launch({ channel: "chrome" })\` / \`launchPersistentContext\`). Capture screenshot + action log. NEVER clear cookies, cache, or site data on the user's live profile. For login state, CLONE the profile first (\`rsync -a <profile>/ <tmp-clone>/\`) and use only the clone as the persistent user-data-dir; clear data only there. |
+| Changes UI rendering | Drive the REAL page from js eval with omowright (staged in the \`browser\` skill): the owned engine (\`connectPipe\` on a task-owned profile, \`connectCloakProfile\` for bot-scored targets) for unauthenticated pages, the attached engine (\`connectBrowserSkill()\` in the user's own signed-in browser) when the page needs their login. Capture screenshot + action log. NEVER clear cookies, cache, or site data on the user's live profile, and never clone it; if the attached engine is missing, run the browser skill's onboarding script and relay its one human step instead of launching a headless browser. |
 | Changes UI rendering or a TUI/terminal layout (incl. CJK/Korean/Japanese/Chinese text) | Load the visual-qa skill: capture reference + actual screenshots (web) or the xterm.js web terminal render (TUI; NEVER \`tmux capture-pane\` - it degrades color and CJK width), run its bundled pixel-diff / column-width script, and get the dual read-only verdict (design-system + functional integrity, and visual fidelity + CJK precision). Record the diff/score artifact. |
 | Changes a desktop/GUI (non-page) surface | Computer use: OS-level GUI automation against the running app. Capture action log + screenshot. |
 | Adds a new tool/hook/feature | Test it end-to-end in a real scenario. |
@@ -103867,36 +103892,29 @@ The real-surface artifact is always required. A test is required only where a co
 **CLEANUP IS PART OF QA \u2014 TRACK IT AS TODOS.** The moment a QA scenario spawns any resource, add a teardown todo for it (QA scripts, tmux assets, browser contexts, PIDs, ports, containers, temp dirs). Execute every teardown todo and capture the receipt before declaring done. A leftover process / tmux session / browser context / bound port / temp dir = NOT done.
 </MANUAL_QA_MANDATE>
 
-### TDD Workflow (MANDATORY on every production code change with a test seam)
-
-Test-first is not optional for code. Every code behavior change \u2014 features, fixes, refactors, perf, glue, config-with-logic \u2014 follows RED \u2192 GREEN \u2192 SURFACE. Prose, docs, and visual-only changes have no seam: skip RED\u2192GREEN and prove them through the surface channel.
-
-1. **RED**: Write the failing test FIRST. Run it. Capture the assertion message proving it fails for the RIGHT reason (not syntax, not import). Paste RED output into the notepad. No production code yet.
-2. **GREEN**: Write the SMALLEST change that flips RED\u2192GREEN. Re-run. Capture GREEN output. If GREEN required ~20+ lines, your test was too coarse \u2014 split it.
-3. **SURFACE**: Exercise the real user-facing surface named by the scenario. Capture artifact path into the notepad.
-4. **REFACTOR**: Optional, only if needed. Tests MUST stay green throughout.
-5. **REGRESSION**: Re-run the FULL scenario list. Record PASS/FAIL inline with both evidence paths.
+### Test Decision (every production code change)
 
-**Refactor exception**: when refactoring behavior whose regressions the change could hide, write characterization tests pinning current observable behavior FIRST, watch them go GREEN against old code, THEN refactor. They remain green throughout.
+1. **READ** the tests covering the area BEFORE touching it \u2014 they are the behavior of record. Note in the notepad: do they encode the intent, cover this path, pass? One WRONG before your change is a FINDING to report \u2014 NEVER edit a test green. A bug: reproduce it first and capture the failure. A refactor: the existing tests are green on the unchanged code first.
+2. **CHANGE**: the SMALLEST change that meets the scenario; update the tests your change makes stale. Add a test ONLY when BOTH hold: the repository keeps tests for this behavior AND a regression would otherwise pass unnoticed by the run and the existing tests \u2014 sized like its neighbors, one case per stated behavior, failing when that behavior breaks. A test that restates the change (a constant, a string, a rename, a call) is NOT evidence; the run is.
+3. **SURFACE**: Exercise the real user-facing surface named by the scenario; a reproduction now passes. Capture the artifact path into the notepad.
+4. **REGRESSION**: Re-run the FULL scenario list plus the step-1 tests. Record PASS/FAIL inline with evidence paths.
 
-**Exemption whitelist** (no new test required): pure formatting, comment-only edits, dependency version bumps with no behavior delta, rename-only moves. Each exemption MUST be justified in \`## Findings\` with the exact reason. Unjustified exemption is rejection.
-
-**If you typed production code without a failing test preceding it in the notepad: STOP, revert, write the test, watch it fail, then redo.**
+Prose, docs, prompt, and visual-only changes have no test seam: review + real-surface QA, NO test \u2014 a test pinning their text is pretend-coverage.
 
 ### Commit Discipline (MANDATORY)
 
-Commit frequently: one atomic commit per verified increment (RED\u2192GREEN + evidence captured), never one end-of-run omnibus. BEFORE composing each message, study the history and mimic it \u2014 run \`git log --oneline -20\` plus \`git log -5 -- <touched paths>\` \u2014 matching subject shape, scope names, message language, body style, and typical commit size. Load the \`git-master\` skill for the commit workflow when available. Skip committing only when the user forbade commits this session.
+Commit frequently: one atomic commit per verified increment (change + evidence captured), never one end-of-run omnibus. BEFORE composing each message, study the history and mimic it \u2014 run \`git log --oneline -20\` plus \`git log -5 -- <touched paths>\` \u2014 matching subject shape, scope names, message language, body style, and typical commit size. Load the \`git-master\` skill for the commit workflow when available. Skip committing only when the user forbade commits this session.
 
 ### Verification Anti-Patterns (BLOCKING)
 
 | Violation | Why It Fails |
 |-----------|--------------|
 | "It should work now" | No evidence. Run it. |
-| "I added the tests" | Did they go RED first, then GREEN? Show both. |
+| "I added the tests" | Did the repository keep tests here, and would the regression have passed unnoticed without them? Otherwise the run is the proof. |
 | "Fixed the bug" | What scenario proves it? Where's the artifact? |
 | "Implementation complete" | Every scenario PASS with both artifacts captured? |
 | Skipping test execution | Tests exist to be RUN, not just written |
-| Writing code before its failing test | TDD floor violated \u2014 revert, write test, redo |
+| A test that restates the change | Cannot fail for any regression \u2014 delete it; the run is the proof |
 
 **CLAIM NOTHING WITHOUT PROOF. EXECUTE. VERIFY. SHOW EVIDENCE.**
 
@@ -104144,43 +104162,37 @@ When the \`create_goal\` tool exists, you MUST register the run's goal with it B
 
 ### SCENARIO CONTRACT (binding, defined BEFORE coding)
 
-Define scenarios sized to the change \u2014 1-2 for a small single-surface change, 3+ for risky or multi-surface work \u2014 each with a binary pass condition and the cheapest faithful proof: a test file+test id at a code seam (test-first), or the real-surface scenario itself when no seam exists (prose, docs, visual-only: review + real-surface QA, no test). Classes:
+Define scenarios sized to the change \u2014 1-2 for a small single-surface change, 3+ for risky or multi-surface work \u2014 each with a binary pass condition and the existing tests that cover it (read first) and the real-surface scenario that proves it (prose, docs, visual-only: review + real-surface QA, no test). Classes:
 - **Happy path** (always)
 - **Edge** (boundary, empty, malformed, concurrent \u2014 when risky)
 - **Adjacent-surface regression** (callers, sibling endpoints, related modules \u2014 when multi-surface)
 
-Scenarios are the contract. Done = every scenario PASSES with both artifacts (RED\u2192GREEN proof AND real-surface artifact).
+Scenarios are the contract. Done = every scenario PASSES with its real-surface artifact captured and the tests of record green.
 
 ### DURABLE NOTEPAD
 
 At start: \`NOTE=$(mktemp -t ulw-$(date +%Y%m%d-%H%M%S).XXXXXX.md)\`. Echo the path. APPEND-ONLY sections: Plan, Scenarios, Now, Todo, Findings (file:line), Learnings. If context is lost, re-read and resume \u2014 this is your only durable memory.
 
-### TDD (MANDATORY for code with a test seam)
-
-Every production code change \u2014 features, fixes, refactors, perf, glue, config-with-logic \u2014 follows RED\u2192GREEN\u2192SURFACE. Prose, docs, and visual-only changes have no seam: skip RED\u2192GREEN, prove them through the surface channel.
-
-1. **RED**: Write the failing test FIRST. Run it. Capture the assertion message that proves it fails for the RIGHT reason (not syntax, not import). Paste RED output into the notepad. No production code yet.
-2. **GREEN**: Smallest change to flip RED\u2192GREEN. Re-run, capture GREEN output. If GREEN required ~20+ lines, your test was too coarse \u2014 split it.
-3. **SURFACE**: Exercise the real user-facing surface (CLI / API / build / UI / config). Capture artifact path.
-4. **REGRESSION**: Re-run the FULL scenario list every increment. Record PASS/FAIL with both artifact paths.
-
-**Refactors**: write characterization tests pinning current observable behavior FIRST, watch them GREEN against the old code, THEN refactor. Stay green throughout.
+### TEST DECISION (every production code change)
 
-**Exemption whitelist**: pure formatting, comment-only edits, version bumps with no behavior delta, rename-only moves. Each MUST be justified in writing. Unjustified exemption = rejection.
+1. **READ** the tests covering the area BEFORE touching it \u2014 the behavior of record: do they encode the intent, cover this path, pass? One WRONG before your change is a FINDING to report \u2014 NEVER edit a test green. A bug: reproduce it first and capture the failure. A refactor: the existing tests are green on the unchanged code first.
+2. **CHANGE**: the SMALLEST change that meets the scenario; update the tests it makes stale. Add a test ONLY when BOTH hold: the repository keeps tests for this behavior AND a regression would otherwise pass unnoticed by the run and the existing tests \u2014 sized like its neighbors, never one that restates the change. The run is the proof, not a test that certifies the diff.
+3. **SURFACE**: Exercise the real user-facing surface (CLI / API / build / UI / config); a reproduction now passes. Capture artifact path.
+4. **REGRESSION**: Re-run the FULL scenario list plus the step-1 tests every increment. Record PASS/FAIL with artifact paths.
 
-**If you typed production code without a failing test preceding it: STOP, revert, write the test, watch it fail, then redo.** No exceptions \u2014 "obvious" / "one-liner" / "too small" do NOT exempt you.
+Prose, docs, and visual-only changes have no seam: prove them through the surface channel, NO test.
 
 ### COMMIT DISCIPLINE (MANDATORY)
 
-Commit frequently: one atomic commit per verified increment (RED\u2192GREEN + evidence captured), never one end-of-run omnibus. BEFORE composing each message, study the history and mimic it \u2014 run \`git log --oneline -20\` plus \`git log -5 -- <touched paths>\` \u2014 matching subject shape, scope names, message language, body style, and typical commit size. Skip committing only when the user forbade commits this session.
+Commit frequently: one atomic commit per verified increment (change + evidence captured), never one end-of-run omnibus. BEFORE composing each message, study the history and mimic it \u2014 run \`git log --oneline -20\` plus \`git log -5 -- <touched paths>\` \u2014 matching subject shape, scope names, message language, body style, and typical commit size. Skip committing only when the user forbade commits this session.
 
 ### Evidence Gates
 
 | Gate | Required Evidence |
 |------|-------------------|
-| **RED** | Failing assertion msg before any production code |
-| **GREEN** | Same test now passing |
-| **Surface** | tmux / curl / browser / Playwright / computer-use / CLI / DB diff artifact path |
+| **Read** | Existing tests for the area noted (intent / coverage / pass) before the change |
+| **Run** | Tests of record green after it; stale expectations updated |
+| **Surface** | tmux / curl / browser (omowright) / computer-use / CLI / DB diff artifact path |
 | **Build** | Exit code 0 |
 | **Suite** | Full run green; no skip/.only/xfail added this turn |
 | **Lint** | lsp_diagnostics clean on changed files |
@@ -104188,13 +104200,13 @@ Commit frequently: one atomic commit per verified increment (RED\u2192GREEN + ev
 <ANTI_OPTIMISM_CHECKPOINT>
 ## BEFORE YOU CLAIM DONE, ANSWER HONESTLY:
 
-1. Did EVERY scenario reach RED captured \u2192 GREEN captured \u2192 surface artifact captured? (paths in notepad)
+1. Did EVERY scenario reach surface artifact captured with the tests of record green? (paths in notepad)
 2. Did I run \`lsp_diagnostics\` and see ZERO errors on changed files? (not "I'm sure")
 3. Did I run the FULL suite and see it PASS? (not "they should pass")
 4. Did I read the actual output of every command? (not skim)
 5. Is EVERY requirement from the request actually implemented? (re-read the request NOW)
 6. Did I classify intent at the start? (if not, my entire approach may be wrong)
-7. Did I write code BEFORE its failing test, anywhere? (if yes, REVERT and redo via TDD)
+7. Did I add a test that only restates the change, or edit a wrong test green? (if yes, delete it / report the finding)
 
 If ANY answer is no \u2192 GO BACK AND DO IT. Do not claim completion.
 </ANTI_OPTIMISM_CHECKPOINT>
@@ -104219,7 +104231,7 @@ Trigger if user said "\uC5C4\uBC00"/"strictly"/"rigorously"/"properly review", o
 | Adds/modifies a CLI command | Run the command with Bash. Show the output. |
 | Changes build output | Run the build. Verify output files exist and are correct. |
 | Modifies API behavior | Call the endpoint. Show the response. |
-| Renders/changes a page | Drive the REAL page from js eval: (1) \`new Bun.WebView()\` on Bun >= 1.4 (macOS default; Linux/Windows need installed Chrome/Chromium/Edge). (2) Otherwise, or for Chrome semantics, stealth, trace, or auth, WRITE a \`playwright-core\` script and run it from the kernel against local Chrome (\`chromium.launch({ channel: "chrome" })\` / \`launchPersistentContext\`). Capture screenshot + action log. NEVER clear cookies, cache, or site data on the user's live profile. For login state, CLONE it first (\`rsync -a <profile>/ <tmp-clone>/\`) and use only the clone as the persistent user-data-dir; clear data only there. |
+| Renders/changes a page | Drive the REAL page from js eval with omowright (staged in the \`browser\` skill): the owned engine (\`connectPipe\` on a task-owned profile, \`connectCloakProfile\` for bot-scored targets) for unauthenticated pages, the attached engine (\`connectBrowserSkill()\` in the user's own signed-in browser) when the page needs their login. Capture screenshot + action log. NEVER clear cookies, cache, or site data on the user's live profile, and never clone it; if the attached engine is missing, run the browser skill's onboarding script and relay its one human step instead of launching a headless browser. |
 | Changes UI rendering or a TUI/terminal layout (incl. CJK/Korean/Japanese/Chinese text) | Load the visual-qa skill: capture reference + actual screenshots (web) or the xterm.js web terminal render (TUI; NEVER \`tmux capture-pane\` - it degrades color and CJK width), run its bundled pixel-diff / column-width script, and get the dual read-only verdict (design-system + functional integrity, and visual fidelity + CJK precision). Record the diff/score artifact. |
 | Drives a desktop/GUI (non-page) surface | Computer use: OS-level GUI automation against the running app. Capture action log + screenshot. |
 | Adds a new tool/hook/feature | Test it end-to-end in a real scenario. |
@@ -104427,17 +104439,14 @@ Before production changes, define scenarios covering:
 
 Each scenario needs a binary pass condition. "Looks good" is not a pass condition.
 
-## TDD WORKFLOW
+## TEST DECISION
 
-TDD is mandatory on production code changes with a test seam; prose, docs, and visual-only changes take review + real-surface QA instead (a test pinning their text is pretend-coverage).
+1. READ: the tests covering the area, before any edit \u2014 intent, coverage, pass. One wrong before your change is a finding, never edited green. A bug: reproduce it first.
+2. CHANGE: the smallest change; update tests it makes stale. A new test only where the repository keeps tests for this behavior AND a regression would otherwise pass unnoticed \u2014 never one that restates the change.
+3. SURFACE: exercise the real user path and capture the artifact; a reproduction now passes.
+4. REGRESSION: rerun the scenario list and the step-1 tests.
 
-1. RED: write or identify a failing test that proves the needed behavior.
-2. GREEN: make the smallest change that flips the test to passing.
-3. SURFACE: exercise the real user path and capture the artifact.
-4. REFACTOR: improve structure only while tests stay green.
-5. REGRESSION: rerun the scenario list.
-
-Exemptions: pure prompt text, formatting, comment-only edits, version bumps with no behavior delta, and rename-only moves. Justify every exemption in the final report.
+Prose, docs, and visual-only changes take review + real-surface QA, no test (a test pinning their text is pretend-coverage).
 
 ## COMMIT DISCIPLINE
 
@@ -104599,8 +104608,8 @@ deep_context = background_output(task_id=...)
 - If delegating: every child prompt carries GOAL, STOP WHEN (the exact observable condition that ends its run \u2014 the child stops the moment it holds), and EVIDENCE (what it returns so you can verify, not trust) \u2014 plus exhaustive context. Judge the child by its returned EVIDENCE against its STOP WHEN, never by its self-report.
 
 **Verify (per-scenario, not just "at the end"):**
-- RED\u2192GREEN proof captured (test id + assertion msg in both states)
-- Real-surface artifact (tmux / curl / browser / Playwright / computer-use / CLI / DB diff)
+- Existing tests read (intent / coverage / pass), stale expectations updated
+- Real-surface artifact (tmux / curl / browser (omowright) / computer-use / CLI / DB diff)
 - \`lsp_diagnostics\` clean on modified files
 - Full suite green, regression scenarios still PASS
 
@@ -104621,17 +104630,13 @@ Maintain a live todo list for every multi-step task: one atomic item per action
 Define scenarios sized to the change \u2014 1-2 for small single-surface work, 3+ for risky or multi-surface work \u2014 covering: **happy path**, plus **edge** (boundary / empty / malformed / concurrent) and **adjacent-surface regression** when the change is risky or multi-surface. For each, write:
 - Binary pass condition ("returns 200 with schema-matching body"), not "should work".
 - The real surface that proves it.
-- The cheapest faithful proof: a test file + test id at a code seam (test-first; see TDD), or the real-surface scenario itself when no seam exists (prose, docs, visual-only: review + real-surface QA, no test).
-
-Scenarios are the contract. Done = every scenario PASSES with RED\u2192GREEN proof AND real-surface artifact captured. Then declare WHEN TO STOP for the whole run, in one line: "I'll stop right away when <the exact observable state that ends this run>" \u2014 its end state MUST be the full STOP GOAL from the Stop rules, never scenario completion alone. The Stop rules bind to this line \u2014 the moment it holds, you stop.
-
-## TDD (MANDATORY on every production code change with a test seam)
+- The existing tests that cover it (see TEST DECISION) and the real-surface scenario that proves it (prose, docs, visual-only: review + real-surface QA, no test).
 
-Code features, fixes, refactors, perf, glue, config-with-logic \u2014 all follow RED\u2192GREEN\u2192SURFACE. Prose, docs, and visual-only changes have no test seam: prove them through the real-surface channel; a test pinning their text is pretend-coverage. Write the failing test FIRST; capture the assertion proving it fails for the right reason; write the SMALLEST change to flip it green; exercise the real surface; capture both artifacts. **If you wrote production code without a failing test preceding it: STOP, revert, write the test, redo.**
+Scenarios are the contract. Done = every scenario PASSES with its real-surface artifact captured and the tests of record green. Then declare WHEN TO STOP for the whole run, in one line: "I'll stop right away when <the exact observable state that ends this run>" \u2014 its end state MUST be the full STOP GOAL from the Stop rules, never scenario completion alone. The Stop rules bind to this line \u2014 the moment it holds, you stop.
 
-Refactors of behavior whose regressions the change could hide: write characterization tests pinning current behavior FIRST, watch them GREEN against old code, THEN refactor. They stay green throughout.
+## TEST DECISION
 
-Exemption whitelist (no new test required): formatting, comment-only, version bumps with no behavior delta, rename-only. Each must be justified in writing. Unjustified exemption is rejection.
+READ the tests covering the area BEFORE touching it \u2014 they are the behavior of record: intent, coverage, pass. One WRONG before your change is a FINDING \u2014 never edit a test green. Reproduce a bug before fixing it; a refactor starts with the existing tests green on the unchanged code. The run proves the change: add a test ONLY where the repository keeps tests for this behavior AND a regression would otherwise pass unnoticed \u2014 sized like its neighbors, never one that restates the change. Prose, docs, and visual-only changes have no test seam: real-surface channel, no test; a test pinning their text is pretend-coverage.
 
 ## COMMIT DISCIPLINE
 
@@ -104641,8 +104646,8 @@ Commit one atomic commit per verified increment; never one end-of-run omnibus. B
 
 | Phase | Action | Required Evidence |
 |-------|--------|-------------------|
-| RED   | Run new test before impl  | Failing assertion with msg |
-| GREEN | Re-run after smallest change | Passing assertion |
+| Read  | Existing tests for the area, before the change | Intent / coverage / pass noted |
+| Run   | Tests of record after the change | Green; stale expectations updated |
 | Surface | Exercise real user path | Artifact path (tmux/curl/browser/...) |
 | Build | Run build command | Exit code 0 |
 | Suite | Full test run | All green; no skip/.only/xfail added |
@@ -104658,7 +104663,7 @@ lsp_diagnostics catches type errors only. Logic bugs, missing behavior, broken f
 | Adds/modifies a CLI command | Run it with Bash. Show output. |
 | Changes build output | Run build. Verify output files. |
 | Modifies API behavior | Call the endpoint. Show response. |
-| Renders/changes a page | Drive the page from js eval: (1) \`new Bun.WebView()\` on Bun >= 1.4 (macOS default; Linux/Windows need installed Chrome/Chromium/Edge). (2) Otherwise, or for Chrome semantics, stealth, trace, or auth, WRITE a \`playwright-core\` script and run it from the kernel against local Chrome (\`chromium.launch({ channel: "chrome" })\` / \`launchPersistentContext\`). Screenshot + action log. NEVER clear cookies, cache, or site data on the user's live profile. For login state, CLONE it first (\`rsync -a <profile>/ <tmp-clone>/\`) and use only the clone as the persistent user-data-dir; clear data only there. |
+| Renders/changes a page | Drive the REAL page from js eval with omowright (staged in the \`browser\` skill): the owned engine (\`connectPipe\` on a task-owned profile, \`connectCloakProfile\` for bot-scored targets) for unauthenticated pages, the attached engine (\`connectBrowserSkill()\` in the user's own signed-in browser) when the page needs their login. Screenshot + action log. NEVER clear cookies, cache, or site data on the user's live profile, and never clone it; if the attached engine is missing, run the browser skill's onboarding script and relay its one human step instead of launching a headless browser. |
 | Changes UI rendering or a TUI/terminal layout (incl. CJK/Korean/Japanese/Chinese text) | Load the visual-qa skill: capture reference + actual screenshots (web) or the xterm.js web terminal render (TUI; NEVER \`tmux capture-pane\` - it degrades color and CJK width), run its bundled pixel-diff / column-width script, and get the dual read-only verdict (design-system + functional integrity, and visual fidelity + CJK precision). Record the diff/score artifact. |
 | Drives a desktop GUI | Computer use: OS-level GUI automation against the running app. Action log + screenshot. |
 | Adds tool/hook/feature | Test end-to-end in a real scenario. |
@@ -104674,7 +104679,7 @@ Trigger if user said "\uC5C4\uBC00"/"strictly"/"rigorously"/"properly review", o
 ## STOP RULES
 
 - After each result, ask whether the user's core request can now be answered with useful evidence in hand. If yes, answer now \u2014 skip any remaining retrieval, ceremony, or verification that adds no evidence.
-- The STOP GOAL: every scenario PASSES with RED\u2192GREEN proof AND real-surface artifact captured; full suite green and \`lsp_diagnostics\` clean on changed files; QA teardown receipts recorded; no scope creep; and (if triggered) the reviewer gate approved unconditionally. Above ALL of that, the decisive test \u2014 outranking every other consideration \u2014 is: is the user's problem ACTUALLY SOLVED in observable behavior? If no, you are NOT done, whatever the checklist says. If yes, deliver the final message and STOP \u2014 no hesitation, no extra verification pass, no polish loop. Work past the stop goal is scope creep, not diligence.
+- The STOP GOAL: every scenario PASSES with its real-surface artifact captured and the tests of record green; full suite green and \`lsp_diagnostics\` clean on changed files; QA teardown receipts recorded; no scope creep; and (if triggered) the reviewer gate approved unconditionally. Above ALL of that, the decisive test \u2014 outranking every other consideration \u2014 is: is the user's problem ACTUALLY SOLVED in observable behavior? If no, you are NOT done, whatever the checklist says. If yes, deliver the final message and STOP \u2014 no hesitation, no extra verification pass, no polish loop. Work past the stop goal is scope creep, not diligence.
 - After 2 identical failed attempts at one step, surface what was tried and ask the user before another retry.
 - After 2 parallel exploration waves yield no new useful facts, stop exploring and act.
 
@@ -132201,6 +132206,10 @@ function resolveCategoryConfig(categoryName, options) {
       return null;
     }
   }
+  if (categoryReq?.requiresAnyModel && availableModels && !hasExplicitUserConfig && !isAnyFallbackModelAvailable(categoryReq.fallbackChain, availableModels)) {
+    log2(`[resolveCategoryConfig] Category ${categoryName} has no available model in its fallback chain`);
+    return null;
+  }
   const defaultPromptAppend = CATEGORY_PROMPT_APPENDS[categoryName] ?? "";
   if (!defaultConfig && !userConfig) {
     return null;
@@ -132391,6 +132400,16 @@ To use this category:
 1. Connect a provider with this model: ${requiredModel}
 2. Or configure an alternative model in your .omo/omo.jsonc for this category
 
+Available categories: ${allCategoryNames}`);
+    }
+    if (categoryExists && requirement?.requiresAnyModel && userCategories?.[categoryName] === undefined) {
+      const chainModels = [...new Set(requirement.fallbackChain.map((entry) => entry.model))].join(", ");
+      return categoryResolutionError(`Category "${categoryName}" has no available model: none of its fallback-chain models (${chainModels}) is available.
+
+To use this category:
+1. Connect a provider that serves one of: ${chainModels}
+2. Or configure a model for this category in your .omo/omo.jsonc
+
 Available categories: ${allCategoryNames}`);
     }
     return categoryResolutionError(`Unknown category: "${categoryName}". Available: ${allCategoryNames}`);

package/dist/skills/browser/ATTRIBUTION.md

@@ -1,25 +1,37 @@
 # ATTRIBUTION / NOTICE
 
-This skill (`browser`, part of `@oh-my-opencode/shared-skills`) is project-original documentation.
-It drives a third-party tool that it does **not** redistribute.
+This skill (`browser`, part of `@oh-my-opencode/shared-skills`) is project-original documentation
+plus a staged copy of one library it drives.
+
+## omowright — bundled
+
+Both engines this skill documents are provided by **omowright**.
+
+- Upstream: the `omowright` repository on GitHub, pinned by commit in the root `package.json`
+  devDependency.
+- License: MIT
+- What ships here: `runtime/omowright/index.js` (the library bundled into one ESM file, including
+  its `ws` and `zod` dependencies) and `runtime/omowright/page-bundle.js` (the in-page snapshot
+  script it reads at import time), staged at build time by
+  `packages/shared-skills/stage-omowright-runtime.mjs` from the pinned root devDependency, with a
+  `manifest.json` recording the version and file digests. The repository holds no copy; the
+  staged files are gitignored and packed through the sibling `.npmignore`.
+- omowright's own license notice is preserved inside the bundled file.
 
 ## BrowserSkill — driven, not vendored
 
-The attached engine this skill documents is **BrowserSkill** by Tencent.
+The attached engine reaches the user's browser through **BrowserSkill** by Tencent.
 
 - Upstream: https://github.com/Tencent/BrowserSkill
 - License: MIT
-- What ships here: documentation and two local helper scripts, authored by this project.
-- What does NOT ship here: the `bsk` CLI, the daemon, the browser extension, or any upstream
-  source. The user installs those from upstream's own installer and from the Chrome Web Store /
-  Edge Add-ons listings. This package never bundles a browser binary or an extension payload.
-
-MIT permits redistribution; we simply have no reason to, because upstream publishes signed
-per-platform releases and store builds that stay current on their own.
+- What ships here: nothing of upstream's. omowright speaks the daemon's IPC protocol; the user
+  installs the `bsk` CLI from upstream's own installer (run by `scripts/browser-install.mjs`) and
+  the extension from the Chrome Web Store / Edge Add-ons listings (registered by the same script
+  through Chrome's external-extension mechanism). This package never bundles a browser binary or
+  an extension payload.
 
 ## Owned-engine references
 
-`references/owned-engine/` documents a *contract* for a code-driven browser — the escalation
-ladder, viewport pinning, reading the network instead of the DOM, overlay and cross-origin
-handling — not any particular package. The techniques are generic CDP practice. No third-party
-source is reproduced.
+`references/owned-engine/` summarizes omowright's own skill and presets (escalation ladder,
+network-first reading, frames, dialogs, human handoff) as routing text; the library's repository
+is the authoritative treatment.