codemini-cli 0.1.18 → 0.2.0

package/README.md

@@ -19,13 +19,13 @@ It is designed for teams that want a coding assistant that feels practical, cont
 
 ### Highlights
 
-- Minimal default tools: `run_command`, `read_file`, `write_file`
+- Minimal default tools: `run`, `read`, `write`
 - Windows-aware shell profile with PowerShell-focused defaults
 - Safe mode enabled by default
 - Built-in lite skills for planning, execution, and collaboration
 - Configurable reply language through `ui.reply_language` (`zh` / `en`)
 - Richer slash completion with priority sorting, inline descriptions, and left/right paging
-- Structured code tools for small models: `locate`, `open_target`, `edit_target`
+- Structured code tools for small models: `grep`, `read`, `edit`
 - More conservative `plan auto` acceptance checks with reviewer/tester goal checklists
 - Tone presets through `soul`, without changing plans or code behavior
 - Example soul presets include `professional`, `playful`, `anime`, `pirate`, `caveman`, and `ceo`
@@ -66,7 +66,7 @@ codemini skill list|install|enable|disable|inspect|reindex
 - Slash completion now prioritizes important commands and config keys, shows short descriptions, and supports `←/→` page switching
 - Ambiguous feature requests can pause for lightweight brainstorming first, and `/brainstorm <question>` gives an explicit way to compare options before coding
 - `plan auto` now turns the original goal into an acceptance checklist, uses a lighter chain only for truly tiny tasks, and treats unmet checklist items as failure signals
-- Structured code tools reduce shell-noise for small models by preferring `locate -> open_target -> edit_target`
+- Structured code tools reduce shell-noise for small models by preferring `grep/read -> edit`
 
 ### Skill Loading
 
@@ -134,13 +134,13 @@ CodeMini CLI 是一个为小模型工作流优化过的代码助手 CLI，重点
 
 ### 主要特点
 
-- 默认工具极简：`run_command`、`read_file`、`write_file`
+- 默认工具极简：`run`、`read`、`write`
 - 面向 Windows 的 PowerShell 默认配置
 - safe mode 默认开启
 - 内置 lite skills，覆盖规划、执行和协作
 - 支持通过 `ui.reply_language` 配置回复语言，当前支持 `zh` / `en`
 - slash 补全支持优先级排序、右侧简短说明和左右分页
-- 为小模型补了结构化代码工具：`locate`、`open_target`、`edit_target`
+- 为小模型补了结构化代码工具：`grep`、`read`、`edit`
 - `plan auto` 会基于原始目标生成验收清单，并更保守地处理 reviewer/tester 结果
 - `soul` 只影响语气，不影响计划和代码行为
 - 可用的 `soul` 示例包括 `professional`、`playful`、`anime`、`pirate`、`caveman`、`ceo`
@@ -181,7 +181,7 @@ codemini skill list|install|enable|disable|inspect|reindex
 - slash 补全会优先展示更重要的命令和配置项，显示简短说明，并支持 `←/→` 翻页
 - 对于需求仍不明确的功能请求，CLI 会先偏向轻量 brainstorm；也可以显式使用 `/brainstorm <问题>` 先比较方案再决定是否编码
 - `plan auto` 会先把原始目标展开成验收清单；只有真正很小的任务才会走轻量链路；如果 reviewer 或 tester 标记了未满足或未验证的验收项，就不会按成功处理
-- 为了减少小模型被 shell 原始输出干扰，新增了 `locate -> open_target -> edit_target` 这套结构化代码工具流
+- 为了减少小模型被 shell 原始输出干扰，新增了 `grep/read -> edit` 这套结构化代码工具流
 
 ### Skill 加载位置
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "codemini-cli",
-  "version": "0.1.18",
+  "version": "0.2.0",
   "description": "Coding CLI optimized for small-model workflows and Windows PowerShell",
   "keywords": [
     "cli",

package/skills/superpowers-lite/SKILL.md

@@ -34,16 +34,17 @@ Routing:
 - do not assume features, storage model, or scope unless the user already gave them
 
 Tool order:
-- prefer `locate` first for repo search and candidate discovery
-- use `open_target` to inspect the smallest useful code block with edit metadata
-- use `edit_target` for minimal validated edits
-- use `search_code`, `read_block`, and `read_symbol_context` when lower-level structured context is needed
+- prefer `grep` first for content search and candidate discovery
+- use `read` to inspect the smallest useful code block
+- use `edit` for minimal focused edits or direct whole-file rewrites when you already have the replacement content
+- use `generate_diff` and `patch` for larger edits or when you already have a diff
+- use `glob` and `list` when you need file or directory discovery
 - use shell search such as `rg` only as a fallback when structured tools are not enough
 
 Core rules:
 
 1. Search first.
-Prefer structured search before broad file reads. Start with `locate`, then inspect with `open_target`, and only fall back to shell search such as `rg` when the structured tools are not enough.
+Prefer structured search before broad file reads. Start with `grep`, then inspect with `read`, and only fall back to shell search such as `rg` when the structured tools are not enough.
 
 2. Keep context tight.
 Do not carry full conversation history into every step. Summarize, narrow scope, and work from the most recent relevant evidence.
@@ -62,18 +63,18 @@ If the requested behavior, scope, or acceptance is unclear, do not jump into imp
 - clear enough to build -> proceed
 
 5. Read and write with intent.
-Use `open_target`, `read_block`, and `read_symbol_context` before `read_file` when possible. Use `edit_target` for focused edits. Use `write_file` only for full-file writes. Avoid unnecessary tool calls and avoid rereading the same file without a reason.
+Use `read` before broad reads when possible. Use `edit` for focused edits or when you already have the complete replacement content. Use `generate_diff` and `patch` for larger changes. Use `write` only for creating new files or explicit whole-file writes. Avoid unnecessary tool calls and avoid rereading the same file without a reason.
 
 6. Verify before claiming success.
 Run the relevant test, check, or command before saying work is fixed or complete.
 
 Default workflow:
-- Search with `locate`
-- Inspect local context with `open_target`
+- Search with `grep`
+- Inspect local context with `read`
 - If the request is unclear, first decide: ask one question, brainstorm, or proceed
 - Plan the next smallest step
 - Delegate if the work is independent
-- Edit with `edit_target`
+- Edit with `edit`
 - Verify
 - Summarize briefly
 

package/src/core/agent-loop.js

@@ -26,11 +26,19 @@ function summarizeToolResult(result) {
       const p = String(obj.path || '');
       const action = String(obj.action || 'write');
       const line = Number(obj.changed_line || 1);
-      const preview = String(obj.diff_preview || '')
-        .split('\n')
-        .slice(0, 3)
-        .join('\n');
-      return `${action} ${p} @L${line}${preview ? `\n${preview}` : ''}`;
+      const suffix =
+        action === 'delete'
+          ? 'deleted'
+          : action === 'create'
+            ? 'created'
+            : action === 'patch'
+              ? 'patched'
+              : action === 'replace_block' || action === 'replace_text'
+                ? 'edited'
+                : action === 'append'
+                  ? 'appended'
+                  : 'updated';
+      return p ? `${suffix} ${p}${line > 0 ? ` @L${line}` : ''}` : suffix;
     }
     if ('path' in obj && 'phase' in obj) {
       const phase = String(obj.phase || '');
@@ -79,6 +87,13 @@ function summarizeToolResult(result) {
       const logs = Array.isArray(obj.recent_logs) ? trimInline(obj.recent_logs.slice(-1)[0] || '', 96) : '';
       return `${taskId || 'service logs'}${logs ? `\n${logs}` : ''}`;
     }
+    if ('files' in obj && Array.isArray(obj.files)) {
+      return `patched ${obj.files.length} file(s)`;
+    }
+    if ('diff' in obj && 'new_hash' in obj && 'path' in obj) {
+      const p = String(obj.path || '');
+      return p ? `diff preview for ${p}` : 'diff preview';
+    }
     if ('created' in obj && Array.isArray(obj.created)) {
       return `created ${obj.created.length} task(s)`;
     }
@@ -98,21 +113,45 @@ function trimInline(value, maxLen = 72) {
   return `${s.slice(0, maxLen - 3)}...`;
 }
 
+function normalizeToolCallName(name) {
+  return String(name || '').trim();
+}
+
 function formatToolDisplayName(name, args) {
-  if (name === 'read_file' || name === 'write_file') {
+  if (name === 'grep') {
+    const query = trimInline(args?.pattern || args?.query || args?.symbol || '', 96);
+    return query ? `grep("${query}")` : 'grep';
+  }
+  if (name === 'glob') {
+    const pattern = trimInline(args?.pattern || '', 96);
+    return pattern ? `glob("${pattern}")` : 'glob';
+  }
+  if (name === 'list') {
+    const target = trimInline(args?.path || '.', 96) || '.';
+    return `list(${target})`;
+  }
+  if (name === 'read' || name === 'write') {
     const target = trimInline(args?.path || '.', 96) || '.';
-    if (name === 'read_file') {
+    if (name === 'read') {
       const start = Number(args?.start_line);
       const end = Number(args?.end_line);
       const hasRange = Number.isFinite(start) && start > 0;
       const suffix = hasRange ? `:${start}-${Number.isFinite(end) && end >= start ? end : start}` : '';
-      return `${name}(${target}${suffix})`;
+      return `read(${target}${suffix})`;
     }
-    return `${name}(${target})`;
+    return `write(${target})`;
   }
-  if (name === 'run_command') {
+  if (name === 'run') {
     const command = trimInline(args?.command || '', 96);
-    return command ? `${name}(${command})` : name;
+    return command ? `run(${command})` : name;
+  }
+  if (name === 'edit') {
+    const target = trimInline(args?.path || args?.file || '.', 96) || '.';
+    return `edit(${target})`;
+  }
+  if (name === 'patch') {
+    const target = trimInline(args?.path || args?.file || args?.patch || '', 96) || '.';
+    return `patch(${target})`;
   }
   if (name === 'start_service') {
     const command = trimInline(args?.command || args?.cmd || '', 96);
@@ -125,20 +164,7 @@ function formatToolDisplayName(name, args) {
     const taskId = trimInline(args?.task_id || args?.taskId || '', 96);
     return taskId ? `${name}(${taskId})` : name;
   }
-  if (
-    name === 'locate' ||
-    name === 'open_target' ||
-    name === 'edit_target' ||
-    name === 'search_code' ||
-    name === 'read_block' ||
-    name === 'read_symbol_context' ||
-    name === 'validate_edit' ||
-    name === 'replace_block' ||
-    name === 'replace_text' ||
-    name === 'insert_before' ||
-    name === 'insert_after' ||
-    name === 'generate_diff'
-  ) {
+  if (name === 'read' || name === 'write' || name === 'run' || name === 'grep' || name === 'glob' || name === 'list' || name === 'edit' || name === 'patch' || name === 'generate_diff') {
     const target = trimInline(args?.path || args?.query || args?.symbol || '', 96);
     return target ? `${name}(${target})` : name;
   }
@@ -218,15 +244,16 @@ export async function runAgentLoop({
 
     for (const call of toolCalls) {
       const args = safeJsonParse(call.arguments);
-      const displayName = formatToolDisplayName(call.name, args);
+      const toolName = normalizeToolCallName(call.name);
+      const displayName = formatToolDisplayName(toolName, args);
       const startedAt = Date.now();
       let approved = true;
-      if (executionMode === 'normal' && !alwaysAllowSet.has(call.name)) {
+      if (executionMode === 'normal' && !alwaysAllowSet.has(toolName)) {
         approved = false;
         if (typeof requestToolApproval === 'function') {
           const decision = await requestToolApproval({
             id: call.id,
-            name: call.name,
+            name: toolName,
             displayName,
             arguments: args
           });
@@ -235,7 +262,7 @@ export async function runAgentLoop({
       }
 
       if (!approved) {
-        if (onEvent) onEvent({ type: 'tool:blocked', name: displayName, id: call.id });
+        if (onEvent) onEvent({ type: 'tool:blocked', name: displayName, id: call.id, arguments: args });
         const blockedMessage = {
           role: 'tool',
           tool_call_id: call.id,
@@ -247,6 +274,7 @@ export async function runAgentLoop({
             type: 'tool:result',
             name: displayName,
             id: call.id,
+            arguments: args,
             content: blockedMessage.content,
             blocked: true
           });
@@ -254,8 +282,8 @@ export async function runAgentLoop({
         continue;
       }
 
-      if (onEvent) onEvent({ type: 'tool:start', name: displayName, id: call.id });
-      const handler = toolHandlers[call.name];
+      if (onEvent) onEvent({ type: 'tool:start', name: displayName, id: call.id, arguments: args });
+      const handler = toolHandlers[toolName];
       if (!handler) {
         throw new Error(`Unknown tool: ${call.name}`);
       }
@@ -270,6 +298,7 @@ export async function runAgentLoop({
             type: 'tool:error',
             name: displayName,
             id: call.id,
+            arguments: args,
             durationMs,
             summary: trimInline(message, 120)
           });
@@ -285,6 +314,7 @@ export async function runAgentLoop({
             type: 'tool:result',
             name: displayName,
             id: call.id,
+            arguments: args,
             content: toolMessage.content,
             error: true
           });
@@ -297,6 +327,7 @@ export async function runAgentLoop({
           type: 'tool:end',
           name: displayName,
           id: call.id,
+          arguments: args,
           durationMs,
           summary: summarizeToolResult(toolResult)
         });
@@ -312,6 +343,7 @@ export async function runAgentLoop({
           type: 'tool:result',
           name: displayName,
           id: call.id,
+          arguments: args,
           content: toolMessage.content
         });
       }

package/src/core/chat-runtime.js

@@ -94,6 +94,8 @@ function describeConfigKey(key, mode = 'set') {
     'shell.default': 'default shell',
     'shell.timeout_ms': 'shell timeout in milliseconds',
     'context.max_tokens': 'context token budget',
+    'soul.preset': 'soul preset',
+    'soul.custom_path': 'custom soul prompt path',
     'policy.safe_mode': 'safe mode switch',
     'policy.allow_dangerous_commands': 'dangerous command allowance'
   };
@@ -1116,12 +1118,28 @@ function effectiveMaxContextTokens(config) {
 }
 
 function buildRuntimeStateSnapshot({ currentSession, config, model, executionMode }) {
-  return {
+  const currentContextTokens = estimateMessagesTokens(currentSession?.messages || []);
+  const maxContextTokens = effectiveMaxContextTokens(config);
+  const contextUsagePct = maxContextTokens > 0 ? Math.min(100, Math.max(0, (currentContextTokens / maxContextTokens) * 100)) : 0;
+  const snapshot = {
     sessionId: currentSession?.id || '',
     mode: executionMode || config.execution?.mode || 'auto',
     model: model || config.model?.name || '',
-    maxContextTokens: effectiveMaxContextTokens(config)
+    maxContextTokens
   };
+  Object.defineProperties(snapshot, {
+    currentContextTokens: {
+      value: currentContextTokens,
+      enumerable: false,
+      writable: false
+    },
+    contextUsagePct: {
+      value: contextUsagePct,
+      enumerable: false,
+      writable: false
+    }
+  });
+  return snapshot;
 }
 
 function estimatePromptTokensForRequest(sessionMessages, userText = '') {
@@ -1345,7 +1363,7 @@ async function askModel({
     onEvent: wrappedAgentEvent,
     executionMode: executionMode || config.execution?.mode || 'auto',
     alwaysAllowTools:
-      alwaysAllowTools || config.execution?.always_allow_tools || ['run_command', 'read_file', 'write_file'],
+      alwaysAllowTools || config.execution?.always_allow_tools || ['run', 'read', 'write'],
     toolResultMaxChars: config.context?.tool_result_max_chars || 12000,
     requestCompletion: async ({ messages, tools, model: selectedModel }) => {
       if (onAgentEvent) onAgentEvent({ type: 'assistant:start' });
@@ -1359,6 +1377,9 @@ async function askModel({
         maxRetries: config.gateway.max_retries ?? 2,
         onTextDelta: (delta) => {
           if (onAgentEvent) onAgentEvent({ type: 'assistant:delta', text: delta });
+        },
+        onToolCallDelta: (toolCall) => {
+          if (onAgentEvent) onAgentEvent({ type: 'assistant:tool_call_delta', toolCall });
         }
       });
     }
@@ -1720,6 +1741,8 @@ export async function createChatRuntime({
     'sessions.retention_days',
     'shell.timeout_ms',
     'context.max_tokens',
+    'soul.preset',
+    'soul.custom_path',
     'policy.safe_mode',
     'policy.allow_dangerous_commands'
   ];
@@ -1851,9 +1874,21 @@ export async function createChatRuntime({
     const body = input.slice(1);
     const tokens = body.trim().split(/\s+/).filter(Boolean);
     const commandPart = tokens[0] || '';
+    const commandHasSubcommands = new Set([
+      'config',
+      'compact',
+      'mode',
+      'tasks',
+      'checkpoint',
+      'plan',
+      'agents',
+      'history',
+      'debug'
+    ]);
 
     const allCommandEntries = listCommandNames();
     const allCommands = allCommandEntries.map((c) => c.name);
+    const exactCommand = Boolean(commandPart) && allCommands.includes(commandPart);
     for (const entry of allCommandEntries) {
       registerSuggestion(`/${entry.name}`, entry.description || '');
     }
@@ -1879,7 +1914,7 @@ export async function createChatRuntime({
       ));
     }
 
-    if (tokens.length === 1 && !hasTrailingSpace) {
+    if (tokens.length === 1 && !hasTrailingSpace && !(exactCommand && commandHasSubcommands.has(commandPart))) {
       const direct = prioritizeByPreferredOrder(
         allCommands
           .filter((name) => name.startsWith(commandPart))
@@ -1891,25 +1926,26 @@ export async function createChatRuntime({
     }
 
     if (commandPart === 'config') {
-      if (tokens.length === 1 || (tokens.length === 2 && !hasTrailingSpace)) {
-        const sub = tokens[1] || '';
+      const subcommand = tokens[1] || '';
+      const subcommandIsExact = ['set', 'get', 'list', 'reset'].includes(subcommand);
+
+      if (tokens.length === 1 || (tokens.length === 2 && !hasTrailingSpace && !subcommandIsExact)) {
         return materializeSuggestions(prioritizeByPreferredOrder(
           ['set', 'get', 'list', 'reset']
-            .filter((s) => s.startsWith(sub))
+            .filter((s) => s.startsWith(subcommand))
             .map((s) => registerSuggestion(`/config ${s}`, configSubcommandDescriptions[`/config ${s}`] || 'config command').value),
           configSubcommandPriority
         ));
       }
 
-      const sub = tokens[1] || '';
-      if (sub === 'get') {
-        const keyPrefix = tokens[2] || '';
+      if (subcommand === 'get') {
+        const keyPrefix = tokens.length >= 3 ? tokens[2] || '' : '';
         return configKeyHints
           .filter((k) => k.startsWith(keyPrefix))
           .map((k) => registerSuggestion(`/config get ${k}`, describeConfigKey(k, 'get')));
       }
-      if (sub === 'set') {
-        const keyPrefix = tokens[2] || '';
+      if (subcommand === 'set') {
+        const keyPrefix = tokens.length >= 3 ? tokens[2] || '' : '';
         return configKeyHints
           .filter((k) => k.startsWith(keyPrefix))
           .map((k) => registerSuggestion(`/config set ${k} `, describeConfigKey(k, 'set')));
@@ -1920,6 +1956,11 @@ export async function createChatRuntime({
 
     if (commandPart === 'compact') {
       const joined = tokens.slice(1).join(' ');
+      if (tokens.length === 1 || (tokens.length === 2 && !hasTrailingSpace)) {
+        return compactOptions
+          .filter((opt) => opt.startsWith(joined) || joined === '')
+          .map((opt) => registerSuggestion(`/compact ${opt}`, 'context compaction command'));
+      }
       return compactOptions
         .filter((opt) => opt.includes(joined) || joined === '')
         .map((opt) => registerSuggestion(`/compact ${opt}`, 'context compaction command'));
@@ -1952,6 +1993,10 @@ export async function createChatRuntime({
     if (commandPart === 'checkpoint') {
       if (tokens.length <= 2 && !hasTrailingSpace) {
         const sub = tokens[1] || '';
+        if (sub === 'list') {
+          return ['--all']
+            .map((v) => registerSuggestion(`/checkpoint list ${v}`, 'checkpoint command'));
+        }
         return ['create', 'list', 'load']
           .filter((s) => s.startsWith(sub))
           .map((s) => registerSuggestion(`/checkpoint ${s}`, 'checkpoint command'));
@@ -1987,6 +2032,10 @@ export async function createChatRuntime({
     if (commandPart === 'agents') {
       if (tokens.length === 1 || (tokens.length === 2 && !hasTrailingSpace)) {
         const sub = tokens[1] || '';
+        if (sub === 'run') {
+          return ['planner', 'coder', 'reviewer', 'tester']
+            .map((r) => registerSuggestion(`/agents run ${r} `, 'sub-agent command'));
+        }
         return ['list', 'run']
           .filter((s) => s.startsWith(sub))
           .map((s) => registerSuggestion(`/agents ${s}`, 'sub-agent command'));
@@ -2001,13 +2050,22 @@ export async function createChatRuntime({
     }
 
     if (commandPart === 'history') {
+      const sub = tokens[1] || '';
       if (tokens.length === 1 || (tokens.length === 2 && !hasTrailingSpace)) {
-        const sub = tokens[1] || '';
+        if (sub === 'resume') {
+          const dynamic = historySessionCache
+            .filter((session) => String(session.id || '').startsWith(''))
+            .map((session) => ({
+              value: `/history resume ${session.id}`,
+              display: `/history resume ${session.id}  ·  ${Number(session.messageCount || 0)} msgs`,
+              description: 'resume a saved session'
+            }));
+          if (dynamic.length > 0) return dynamic;
+        }
         return ['list', 'current', 'resume']
           .filter((s) => s.startsWith(sub))
-          .map((s) => `/history ${s}`);
+          .map((s) => registerSuggestion(`/history ${s}`, 'history command'));
       }
-      const sub = tokens[1] || '';
       if (sub === 'resume') {
         const idPrefix = tokens[2] || '';
         const dynamic = historySessionCache
@@ -2025,7 +2083,15 @@ export async function createChatRuntime({
 
     if (commandPart === 'debug') {
       const sub = tokens[1] || '';
-      if (!sub) return debugTemplates;
+      if (tokens.length === 1 || (tokens.length === 2 && !hasTrailingSpace)) {
+        if (sub === 'keys') {
+          return ['on', 'off', 'status']
+            .map((v) => registerSuggestion(`/debug keys ${v}`, 'keyboard debug command'));
+        }
+        return ['keys']
+          .filter((s) => s.startsWith(sub))
+          .map((s) => registerSuggestion(`/debug ${s}`, 'debug command'));
+      }
       if (sub === 'keys') {
         const action = tokens[2] || '';
         return ['on', 'off', 'status']
@@ -2609,6 +2675,13 @@ export async function createChatRuntime({
     }
 
     const expandedText = await expandFileMentions(parsedInput.text, process.cwd());
+    const selectedAutoSkills = selectAutoSkillNames(expandedText).filter((name) => isSkillEnabled(config, name));
+    if (selectedAutoSkills.length > 0 && onAgentEvent) {
+      onAgentEvent({
+        type: 'skill:auto',
+        names: selectedAutoSkills
+      });
+    }
     const routedSystemPrompt = buildAutoSkillSystemPrompt(activeReplySystemPrompt, commands, config, expandedText);
     const result = await askModel({
       text: expandedText,

package/src/core/command-policy.js

@@ -16,13 +16,13 @@ function suggestionForToken(token, config) {
   const shell = String(config?.shell?.default || '').toLowerCase();
   if (token === 'find' || token === 'grep') {
     return shell === 'powershell'
-      ? 'Prefer structured tools like locate and open_target first. If you need shell fallback, use allowed search and context commands such as Get-ChildItem, Select-String, Get-Content, or rg when available.'
-      : 'Prefer structured tools like locate and open_target first. If you need shell fallback, use allowed search and context commands such as rg, find, grep, sed, cat, or ls.';
+      ? 'Prefer structured tools like grep, list, read, and edit first. If you need shell fallback, use allowed search and context commands such as Get-ChildItem, Select-String, Get-Content, or rg when available.'
+      : 'Prefer structured tools like grep, glob, list, read, and edit first. If you need shell fallback, use allowed search and context commands such as rg, find, grep, sed, cat, or ls.';
   }
   if (shell === 'powershell') {
-    return 'Prefer structured tools like locate, open_target, and edit_target first. If you need shell fallback, use allowed shell commands for search and local context such as Get-ChildItem, Get-Content, Select-String, or rg when available.';
+    return 'Prefer structured tools like read, edit, write, grep, and list first. If you need shell fallback, use allowed shell commands for search and local context such as Get-ChildItem, Get-Content, Select-String, or rg when available.';
   }
-  return 'Prefer structured tools like locate, open_target, and edit_target first. If you need shell fallback, use allowed shell commands for search and local context such as rg, find, grep, sed, cat, or ls.';
+  return 'Prefer structured tools like read, edit, write, grep, glob, and list first. If you need shell fallback, use allowed shell commands for search and local context such as rg, find, grep, sed, cat, or ls.';
 }
 
 export function evaluateCommandPolicy(command, config, workspaceRoot = process.cwd()) {

package/src/core/config-store.js

@@ -34,26 +34,20 @@ const DEFAULT_CONFIG = {
   execution: {
     mode: 'auto',
     always_allow_tools: [
-      'locate',
-      'open_target',
-      'edit_target',
-      'search_code',
-      'read_block',
-      'read_symbol_context',
-      'validate_edit',
-      'replace_block',
-      'replace_text',
-      'insert_before',
-      'insert_after',
+      'read',
+      'grep',
+      'glob',
+      'list',
+      'edit',
+      'write',
+      'run',
+      'patch',
       'generate_diff',
       'start_service',
       'list_services',
       'get_service_status',
       'get_service_logs',
-      'stop_service',
-      'run_command',
-      'read_file',
-      'write_file'
+      'stop_service'
     ],
     max_steps: 16
   },
@@ -134,26 +128,19 @@ function normalizePolicyLists(config) {
     : [];
   next.execution.always_allow_tools = uniqueStrings(
     [
-      'locate',
-      'open_target',
-      'edit_target',
-      'search_code',
-      'read_block',
-      'read_symbol_context',
-      'validate_edit',
-      'replace_block',
-      'replace_text',
-      'insert_before',
-      'insert_after',
+      'read',
+      'grep',
+      'glob',
+      'list',
+      'edit',
+      'write',
+      'run',
       'generate_diff',
       'start_service',
       'list_services',
       'get_service_status',
       'get_service_logs',
       'stop_service',
-      'run_command',
-      'read_file',
-      'write_file',
       ...rawTools
     ].filter((name) => String(name) !== 'list_files')
   );

package/src/core/provider/openai-compatible.js

@@ -205,6 +205,7 @@ export async function createChatCompletionStream({
   temperature = 0.2,
   tools,
   onTextDelta,
+  onToolCallDelta,
   timeoutMs = 90000,
   maxRetries = 2
 }) {
@@ -248,6 +249,14 @@ export async function createChatCompletionStream({
       if (td.function?.name) current.name = `${current.name}${td.function.name}`;
       if (td.function?.arguments) current.arguments = `${current.arguments}${td.function.arguments}`;
       toolCallsByIndex.set(idx, current);
+      if (onToolCallDelta) {
+        onToolCallDelta({
+          index: idx,
+          id: current.id || `tc-${idx + 1}`,
+          name: current.name,
+          arguments: current.arguments || '{}'
+        });
+      }
     }
   }
 

package/src/core/shell-profile.js

@@ -118,5 +118,5 @@ export function getEffectivePolicy(config) {
 
 export function getShellSystemPrompt(value) {
   const profile = getShellProfile(value);
-  return `You are CodeMini CLI working in a ${profile.label} shell environment. Prefer the high-level structured workflow first: use locate to find candidates, open_target to inspect the smallest useful block and receive edit metadata, and edit_target to apply minimal edits. When you need lower-level control, use search_code, read_block, read_symbol_context, validate_edit, replace_block, replace_text, insert_before, insert_after, and generate_diff. Use start_service, list_services, get_service_status, get_service_logs, and stop_service for long-running servers or watchers. Use run_command only for one-shot commands that should exit on their own. Use read_file only when structured reads are not enough. Use write_file only for full-file writes and always provide a concrete file path, not a directory. For existing code files, prefer locate -> open_target -> edit_target and only use write_file with full_file_rewrite=true when a whole-file rewrite is truly intended. Avoid unnecessary tool calls.`;
+  return `You are CodeMini CLI working in a ${profile.label} shell environment. Prefer OpenCode-style primary tools first: use read to inspect files, grep to search file contents, glob to find files by pattern, list to inspect directories, edit to modify existing files, write to create or fully rewrite files when appropriate, patch to apply unified diffs, and run for one-shot shell commands like install, build, test, or other finite tasks. Classify frontend, backend, database, and Docker work carefully: use run for finite commands, and use start_service, list_services, get_service_status, get_service_logs, and stop_service for long-running servers, watchers, and dev processes. Treat edit as the default editing path for existing code. Internal low-level edit strategies such as target resolution, block replacement, exact text replacement, and anchored inserts are handled inside edit rather than exposed as separate tools. Use generate_diff when you need a structured preview of a proposed file change. For existing code files, prefer grep/read/edit and only use write with full_file_rewrite=true when a whole-file rewrite is truly intended. Avoid unnecessary tool calls.`;
 }