Merge remote-tracking branch 'origin/main' into feat/migrate-review-agents

package.json

@@ -77,7 +77,8 @@
     "@agentclientprotocol/sdk": "^0.19.0",
     "@claude-code-best/mcp-chrome-bridge": "^2.0.8",
     "ws": "^8.20.0",
-    "gray-matter": "^4.0.3"
+    "gray-matter": "^4.0.3",
+    "undici": "^7.24.6"
   },
   "devDependencies": {
     "@alcalzone/ansi-tokenize": "^0.3.0",

src/bootstrap/state.ts

@@ -195,6 +195,8 @@ type State = {
   sdkBetas: string[] | undefined
   // Main thread agent type (from --agent flag or settings)
   mainThreadAgentType: string | undefined
+  // Currently active skill name (set when a /skill-name is invoked)
+  activeSkillName: string | undefined
   // Remote mode (--remote flag)
   isRemoteMode: boolean
   // Direct connect server URL (for display in header)
@@ -381,6 +383,8 @@ function getInitialState(): State {
     sdkBetas: undefined,
     // Main thread agent type
     mainThreadAgentType: undefined,
+    // Currently active skill name
+    activeSkillName: undefined,
     // Remote mode
     isRemoteMode: false,
     ...(process.env.USER_TYPE === 'ant'
@@ -1622,6 +1626,14 @@ export function setMainThreadAgentType(agentType: string | undefined): void {
   STATE.mainThreadAgentType = agentType
 }
 
+export function getActiveSkillName(): string | undefined {
+  return STATE.activeSkillName
+}
+
+export function setActiveSkillName(skillName: string | undefined): void {
+  STATE.activeSkillName = skillName
+}
+
 export function getIsRemoteMode(): boolean {
   return STATE.isRemoteMode
 }

src/commands/clear/caches.ts

@@ -5,6 +5,7 @@
 import { feature } from 'bun:bundle'
 import {
   clearInvokedSkills,
+  setActiveSkillName,
   setLastEmittedDate,
 } from '../../bootstrap/state.js'
 import { clearCommandsCache } from '../../commands.js'
@@ -115,6 +116,8 @@ export function clearSessionCaches(
   if (!hasPreserved) clearAllDumpState()
   // Clear invoked skills cache (each entry holds full skill file content)
   clearInvokedSkills(preservedAgentIds)
+  // Reset active skill name so new session doesn't inherit previous session's agent-type
+  setActiveSkillName(undefined)
   // Clear git dir resolution cache
   clearResolveGitDirCache()
   // Clear dynamic skills (loaded from skill directories)

src/constants/system.ts

@@ -1,11 +1,6 @@
 // Critical system constants extracted to break circular dependencies
 
-import { feature } from 'bun:bundle'
-import { getFeatureValue_CACHED_MAY_BE_STALE } from '../services/analytics/growthbook.js'
-import { logForDebugging } from '../utils/debug.js'
-import { isEnvDefinedFalsy } from '../utils/envUtils.js'
 import { getAPIProvider } from '../utils/model/providers.js'
-import { getWorkload } from '../utils/workloadContext.js'
 
 const DEFAULT_PREFIX = `You are Claude Code, Anthropic's official CLI for Claude.`
 const AGENT_SDK_CLAUDE_CODE_PRESET_PREFIX = `You are Claude Code, Anthropic's official CLI for Claude, running within the Claude Agent SDK.`
@@ -45,51 +40,6 @@ export function getCLISyspromptPrefix(options?: {
   return DEFAULT_PREFIX
 }
 
-/**
+export function getAttributionHeader(_fingerprint: string): string {
- * Check if attribution header is enabled.
+  return ''
- * Enabled by default, can be disabled via env var or GrowthBook killswitch.
- */
-function isAttributionHeaderEnabled(): boolean {
-  if (isEnvDefinedFalsy(process.env.CLAUDE_CODE_ATTRIBUTION_HEADER)) {
-    return false
-  }
-  return getFeatureValue_CACHED_MAY_BE_STALE('tengu_attribution_header', true)
-}
-
-/**
- * Get attribution header for API requests.
- * Returns a header string with cc_version (including fingerprint) and cc_entrypoint.
- * Enabled by default, can be disabled via env var or GrowthBook killswitch.
- *
- * When NATIVE_CLIENT_ATTESTATION is enabled, includes a `cch=00000` placeholder.
- * Before the request is sent, Bun's native HTTP stack finds this placeholder
- * in the request body and overwrites the zeros with a computed hash. The
- * server verifies this token to confirm the request came from a real Claude
- * Code client. See bun-anthropic/src/http/Attestation.zig for implementation.
- *
- * We use a placeholder (instead of injecting from Zig) because same-length
- * replacement avoids Content-Length changes and buffer reallocation.
- */
-export function getAttributionHeader(fingerprint: string): string {
-  if (!isAttributionHeaderEnabled()) {
-    return ''
-  }
-
-  const version = `${MACRO.VERSION}.${fingerprint}`
-  const entrypoint = process.env.CLAUDE_CODE_ENTRYPOINT ?? 'unknown'
-
-  // cch=00000 placeholder is overwritten by Bun's HTTP stack with attestation token
-  const cch = feature('NATIVE_CLIENT_ATTESTATION') ? ' cch=00000;' : ''
-  // cc_workload: turn-scoped hint so the API can route e.g. cron-initiated
-  // requests to a lower QoS pool. Absent = interactive default. Safe re:
-  // fingerprint (computed from msg chars + version only, line 78 above) and
-  // cch attestation (placeholder overwritten in serialized body bytes after
-  // this string is built). Server _parse_cc_header tolerates unknown extra
-  // fields so old API deploys silently ignore this.
-  const workload = getWorkload()
-  const workloadPair = workload ? ` cc_workload=${workload};` : ''
-  const header = `x-anthropic-billing-header: cc_version=${version}; cc_entrypoint=${entrypoint};${cch}${workloadPair}`
-
-  logForDebugging(`attribution header ${header}`)
-  return header
 }

src/costrict/provider/fetch.ts

@@ -38,6 +38,15 @@ type CoStrictFetch = typeof fetch & {
   preconnect?: (url: string | URL) => void
 }
 
+// PascalCase / camelCase → kebab-case: "StrictSpec" → "strict-spec", "TDD" → "tdd"
+function toKebabCase(s: string | undefined): string | undefined {
+  if (!s) return undefined
+  return s
+    .replace(/([A-Z]+)([A-Z][a-z])/g, '$1-$2')
+    .replace(/([a-z\d])([A-Z])/g, '$1-$2')
+    .toLowerCase()
+}
+
 /**
  * 创建自定义 fetch 函数，用于 CoStrict API 请求
  *
@@ -47,7 +56,10 @@ type CoStrictFetch = typeof fetch & {
  * 3. 注入 Authorization 和 CoStrict 特有 headers
  * 4. 反应性 401 错误恢复（自动重试一次）
  */
-export function createCoStrictFetch(): CoStrictFetch {
+export function createCoStrictFetch(options?: {
+  agentType?: string
+}): CoStrictFetch {
+  const agentType = toKebabCase(options?.agentType) || 'build'
   const costrictFetch = async (
     input: RequestInfo | URL,
     init?: RequestInit,
@@ -94,6 +106,7 @@ export function createCoStrictFetch(): CoStrictFetch {
     headers.set('X-Title', 'CoStrict-CLI')
     headers.set('X-Costrict-Version', `costrict-cli-${VERSION}`)
     headers.set('X-Request-ID', randomUUID())
+    headers.set('agent-type', agentType)
     headers.set('zgsm-client-id', creds.machine_id)
     headers.set('zgsm-client-ide', 'cli')
 

src/costrict/provider/index.ts

@@ -33,6 +33,8 @@ import { resolveCoStrictModel } from './modelMapping.js'
 import { getCoStrictBaseURL } from './auth.js'
 import { loadCoStrictCredentials } from './credentials.js'
 import { isOpenAIThinkingEnabled } from '../../services/api/openai/requestBody.js'
+import { fetchCoStrictModels } from './models.js'
+import { getMainThreadAgentType, getActiveSkillName } from '../../bootstrap/state.js'
 
 /**
  * CoStrict 查询路径
@@ -57,10 +59,28 @@ export async function* queryModelCoStrict(
     const baseUrl = getCoStrictBaseURL(creds?.base_url)
     const chatBaseURL = `${baseUrl}/chat-rag/api/v1`
 
-    // 3. 规范化消息
+    // 3. 从模型列表获取 maxTokens 相关参数
+    let maxTokensParamKey: string = 'max_tokens'
+    let maxTokensValue: number | undefined
+    if (creds?.access_token) {
+      try {
+        const modelList = await fetchCoStrictModels(baseUrl, creds.access_token)
+        const modelInfo = modelList.find(m => m.id === costrictModel)
+        if (modelInfo) {
+          maxTokensParamKey = modelInfo.maxTokensKey || 'max_tokens'
+          if (modelInfo.maxTokens != null) {
+            maxTokensValue = modelInfo.maxTokens
+          }
+        }
+      } catch {
+        // 获取模型列表失败，使用默认值
+      }
+    }
+
+    // 4. 规范化消息
     const messagesForAPI = normalizeMessagesForAPI(messages, tools)
 
-    // 4. 构建工具 schema
+    // 5. 构建工具 schema
     const toolSchemas = await Promise.all(
       tools.map(tool =>
         toolToAPISchema(tool, {
@@ -81,7 +101,7 @@ export async function* queryModelCoStrict(
       },
     )
 
-    // 5. 转换为 OpenAI 格式
+    // 6. 转换为 OpenAI 格式
     // 根据模型名称自动检测是否启用thinking模式
     const enableThinking = isOpenAIThinkingEnabled(costrictModel)
     const openaiMessages = anthropicMessagesToOpenAI(
@@ -92,8 +112,10 @@ export async function* queryModelCoStrict(
     const openaiTools = anthropicToolsToOpenAI(standardTools)
     const openaiToolChoice = anthropicToolChoiceToOpenAI(options.toolChoice)
 
-    // 6. 创建专用的 CoStrict OpenAI 客户端（不缓存，每次使用新的 fetch）
+    // 7. 创建专用的 CoStrict OpenAI 客户端（不缓存，每次使用新的 fetch）
-    const costrictFetch = createCoStrictFetch()
+    const costrictFetch = createCoStrictFetch({
+      agentType: getMainThreadAgentType() ?? getActiveSkillName(),
+    })
     const client = new OpenAI({
       apiKey: 'costrict-managed', // 实际 token 由 createCoStrictFetch 注入
       baseURL: chatBaseURL,
@@ -111,28 +133,32 @@ export async function* queryModelCoStrict(
       `[CoStrict] model=${costrictModel}, baseURL=${chatBaseURL}, messages=${openaiMessages.length}, tools=${openaiTools.length}`,
     )
 
-    // 7. 调用 API（流式）
+    // 8. 调用 API（流式）
+    const requestBody: Record<string, unknown> = {
+      model: costrictModel,
+      messages: openaiMessages,
+      ...(openaiTools.length > 0 && {
+        tools: openaiTools,
+        ...(openaiToolChoice && {
+          tool_choice:
+            openaiToolChoice as OpenAI.Chat.Completions.ChatCompletionToolChoiceOption,
+        }),
+      }),
+      stream: true,
+      stream_options: { include_usage: true },
+      ...(options.temperatureOverride !== undefined && {
+        temperature: options.temperatureOverride,
+      }),
+      ...(maxTokensValue != null && {
+        [maxTokensParamKey]: maxTokensValue,
+      }),
+    }
     const stream = await client.chat.completions.create(
-      {
+      requestBody as unknown as OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming,
-        model: costrictModel,
-        messages: openaiMessages,
-        ...(openaiTools.length > 0 && {
-          tools: openaiTools,
-          ...(openaiToolChoice && {
-            tool_choice:
-              openaiToolChoice as OpenAI.Chat.Completions.ChatCompletionToolChoiceOption,
-          }),
-        }),
-        stream: true,
-        stream_options: { include_usage: true },
-        ...(options.temperatureOverride !== undefined && {
-          temperature: options.temperatureOverride,
-        }),
-      },
       { signal },
     )
 
-    // 8. 转换流并 yield 事件
+    // 9. 转换流并 yield 事件
     const adaptedStream = adaptOpenAIStreamToAnthropic(stream, costrictModel)
 
     const contentBlocks: Record<number, any> = {}

src/costrict/provider/models.ts

@@ -12,6 +12,7 @@ export interface CoStrictModel {
   supportsImages?: boolean
   contextWindow?: number
   maxTokens?: number
+  maxTokensKey?: string
   creditConsumption?: number
   creditDiscount?: number
   [key: string]: any
@@ -42,6 +43,7 @@ export async function fetchCoStrictModels(
       headers: {
         Authorization: `Bearer ${accessToken}`,
         Accept: 'application/json',
+        'User-Agent': `csc/${MACRO.VERSION}`,
       },
     })
 

src/utils/processUserInput/processSlashCommand.tsx

@@ -25,7 +25,7 @@ import type {
   ProgressMessage,
   UserMessage,
 } from 'src/types/message.js'
-import { addInvokedSkill, getSessionId } from '../../bootstrap/state.js'
+import { addInvokedSkill, getSessionId, setActiveSkillName } from '../../bootstrap/state.js'
 import { COMMAND_MESSAGE_TAG, COMMAND_NAME_TAG } from '../../constants/xml.js'
 import type { CanUseToolFn } from '../../hooks/useCanUseTool.js'
 import {
@@ -162,6 +162,8 @@ async function executeForkedSlashCommand(
       ? { ...baseAgent, effort: command.effort }
       : baseAgent
 
+  setActiveSkillName(agentDefinition.agentType ?? command.name)
+
   logForDebugging(
     `Executing forked slash command /${command.name} with agent ${agentDefinition.agentType}`,
   )
@@ -1230,6 +1232,7 @@ async function getMessagesForPromptSlashCommand(
     skillContent,
     getAgentContext()?.agentId ?? null,
   )
+  setActiveSkillName(command.name)
 
   const metadata = formatCommandLoadingMetadata(command, args)