feat: 品牌替换 + 启动优化 + AGENTS.md 模板定制

- 品牌替换：OpenCode/opencode → AirCoding/aircoding（16+ 文件）
- Logo ASCII art：修复 left/right 行数不匹配导致的启动崩溃
- 启动诊断：添加 OPENCODE_PRINT_TIMING 计时探针
- dev 模式默认 --pure 跳过外部插件加载
- AGENTS.md 模板：追加 AirCoding 多 Agent 专项段落
- architect prompt + plugin：强化 AGENTS.md 产出验证

packages/opencode/specs/v2/api.ts

@@ -0,0 +1,67 @@
+// @ts-nocheck
+
+import { OpenCode } from "@opencode-ai/core"
+import { ReadTool } from "@opencode-ai/core/tools"
+
+const opencode = OpenCode.make({})
+
+opencode.tool.add(ReadTool)
+
+opencode.tool.add({
+  name: "bash",
+  schema: {
+    type: "object",
+    properties: {
+      command: {
+        type: "string",
+        description: "The command to run.",
+      },
+    },
+    required: ["command"],
+  },
+  execute(input, ctx) {},
+})
+
+opencode.auth.add({
+  provider: "openai",
+  type: "api",
+  value: process.env.OPENAI_API_KEY,
+})
+
+opencode.agent.add({
+  name: "build",
+  permissions: [],
+  model: {
+    id: "gpt-5-5",
+    provider: "openai",
+    variant: "xhigh",
+  },
+})
+
+const sessionID = await opencode.session.create({
+  agent: "build",
+})
+
+opencode.subscribe((event) => {
+  console.log(event)
+})
+
+await opencode.session.prompt({
+  sessionID,
+  text: "hey what is up",
+})
+
+await opencode.session.prompt({
+  sessionID,
+  text: "what is up with this",
+  files: [
+    {
+      mime: "image/png",
+      uri: "data:image/png;base64,xxxx",
+    },
+  ],
+})
+
+await opencode.session.wait()
+
+console.log(await opencode.session.messages(sessionID))

packages/opencode/specs/v2/message-shape.md

@@ -0,0 +1,136 @@
+# Message Shape
+
+Problem:
+
+- stored messages need enough data to replay and resume a session later
+- prompt hooks often just want to append a synthetic user/assistant message
+- today that means faking ids, timestamps, and request metadata
+
+## Option 1: Two Message Shapes
+
+Keep `User` / `Assistant` for stored history, but clean them up.
+
+```ts
+type User = {
+  role: "user"
+  time: { created: number }
+  request: {
+    agent: string
+    model: ModelRef
+    variant?: string
+    format?: OutputFormat
+    system?: string
+    tools?: Record<string, boolean>
+  }
+}
+
+type Assistant = {
+  role: "assistant"
+  run: { agent: string; model: ModelRef; path: { cwd: string; root: string } }
+  usage: { cost: number; tokens: Tokens }
+  result: { finish?: string; error?: Error; structured?: unknown; kind: "reply" | "summary" }
+}
+```
+
+Add a separate transient `PromptMessage` for prompt surgery.
+
+```ts
+type PromptMessage = {
+  role: "user" | "assistant"
+  parts: PromptPart[]
+}
+```
+
+Plugin hook example:
+
+```ts
+prompt.push({
+  role: "user",
+  parts: [{ type: "text", text: "Summarize the tool output above and continue." }],
+})
+```
+
+Tradeoff: prompt hooks get easy lightweight messages, but there are now two message shapes.
+
+## Option 2: Prompt Mutators
+
+Keep `User` / `Assistant` as the stored history model.
+
+Prompt hooks do not build messages directly. The runtime gives them prompt mutators.
+
+```ts
+type PromptEditor = {
+  append(input: { role: "user" | "assistant"; parts: PromptPart[] }): void
+  prepend(input: { role: "user" | "assistant"; parts: PromptPart[] }): void
+  appendTo(target: "last-user" | "last-assistant", parts: PromptPart[]): void
+  insertAfter(messageID: string, input: { role: "user" | "assistant"; parts: PromptPart[] }): void
+  insertBefore(messageID: string, input: { role: "user" | "assistant"; parts: PromptPart[] }): void
+}
+```
+
+Plugin hook examples:
+
+```ts
+prompt.append({
+  role: "user",
+  parts: [{ type: "text", text: "Summarize the tool output above and continue." }],
+})
+```
+
+```ts
+prompt.appendTo("last-user", [{ type: "text", text: BUILD_SWITCH }])
+```
+
+Tradeoff: avoids a second full message type and avoids fake ids/timestamps, but moves more magic into the hook API.
+
+## Option 3: Separate Turn State
+
+Move execution settings out of `User` and into a separate turn/request object.
+
+```ts
+type Turn = {
+  id: string
+  request: {
+    agent: string
+    model: ModelRef
+    variant?: string
+    format?: OutputFormat
+    system?: string
+    tools?: Record<string, boolean>
+  }
+}
+
+type User = {
+  role: "user"
+  turnID: string
+  time: { created: number }
+}
+
+type Assistant = {
+  role: "assistant"
+  turnID: string
+  usage: { cost: number; tokens: Tokens }
+  result: { finish?: string; error?: Error; structured?: unknown; kind: "reply" | "summary" }
+}
+```
+
+Examples:
+
+```ts
+const turn = {
+  request: {
+    agent: "build",
+    model: { providerID: "openai", modelID: "gpt-5" },
+  },
+}
+```
+
+```ts
+const msg = {
+  role: "user",
+  turnID: turn.id,
+  parts: [{ type: "text", text: "Summarize the tool output above and continue." }],
+}
+```
+
+Tradeoff: stored messages get much smaller and cleaner, but replay now has to join messages with turn state and prompt hooks still need a way to pick which turn they belong to.

packages/opencode/specs/v2/notifications.md

@@ -0,0 +1,13 @@
+# TUI Notifications Default
+
+Problem:
+
+- v1 defaults `attention.enabled` to `false`
+- users can opt in with `attention.enabled = true`
+- v2 should make core TUI notifications a default behavior
+
+## v2 Target
+
+Flip `attention.enabled` to `true` by default in v2.
+
+Keep `attention.enabled = false` as the explicit opt-out.

packages/opencode/specs/v2/tui-command-shim.md

@@ -0,0 +1,67 @@
+# TUI Command Shim Removal
+
+Problem:
+
+- v1 keeps a deprecated `api.command` TUI plugin shim so older plugins do not fail during initialization
+- v2 should expose only the keymap command API
+- tests and fixtures should not encode legacy command behavior as expected behavior
+
+## Remove Public Types
+
+In `packages/plugin/src/tui.ts`, remove:
+
+- `TuiCommand`
+- `TuiCommandApi`
+- `TuiPluginApi.command`
+
+Keep `api.keymap` as the only TUI command registration and execution surface.
+
+## Remove Runtime Shim
+
+Delete `packages/opencode/src/cli/cmd/tui/plugin/command-shim.ts`.
+
+In `packages/opencode/src/cli/cmd/tui/plugin/api.tsx`, remove:
+
+- the `createCommandShim` import
+- the `command: createCommandShim(...)` field from `createTuiApi(...)`
+
+In `packages/opencode/src/cli/cmd/tui/plugin/runtime.ts`, remove:
+
+- the `createCommandShim` import
+- the `command: createCommandShim(...)` field from `pluginApi(...)`
+
+## Migration Target
+
+Plugin authors should replace old calls with keymap calls:
+
+```ts
+api.keymap.registerLayer({
+  commands: [
+    {
+      name: "plugin.command",
+      title: "Plugin Command",
+      namespace: "palette",
+      slashName: "plugin",
+      run() {
+        api.ui.dialog.clear()
+      },
+    },
+  ],
+  bindings: [{ key: "ctrl+shift+p", cmd: "plugin.command" }],
+})
+```
+
+Direct replacements:
+
+- `api.command.register(cb)` -> `api.keymap.registerLayer({ commands, bindings })`
+- `api.command.trigger(name)` -> `api.keymap.dispatchCommand(name)`
+- `api.command.show()` -> `api.keymap.dispatchCommand("command.palette.show")`
+- `onSelect(dialog)` -> use `api.ui.dialog` from the plugin API closure
+
+## Verification
+
+After removal, run from package directories:
+
+- `bun typecheck` in `packages/plugin`
+- `bun typecheck` in `packages/opencode`
+- TUI plugin loader tests in `packages/opencode` if runtime plugin API wiring changed