natureco-cli 5.63.0 → 5.64.1

package/CHANGELOG.md

@@ -2,6 +2,43 @@
 
 All notable changes to NatureCo CLI will be documented in this file.
 
+## [5.64.1] - 2026-07-13 — Reliable coding context and token-budgeted follow-ups
+
+### Fixed
+- `natureco code` now passes prior user/assistant turns into each workflow call, so follow-up requests correctly refer to files and work created earlier in the same session.
+- Workflow history excludes system/tool internals, empty messages and stale turns; oversized recent responses are safely truncated instead of being resent in full.
+- The help screen now parses provider URLs correctly (`api.minimax.io` instead of `https:`).
+
+### Token economy
+- Added workflow-history budgets to all profiles: Efficient 1,024 tokens, Balanced 2,048 tokens and Quality 8,192 tokens.
+- Long generated HTML/code no longer dominates every later prompt. A synthetic 32,000-character response is bounded from roughly 8,000 repeated tokens to 2,048 in Balanced mode (approximately 74% less repeated context).
+- Preserved recent-turn continuity while keeping token use bounded by both message count and estimated tokens.
+
+### Documentation and tests
+- Updated README release highlights, tool count, coding-session behavior and token-economy documentation.
+- Added regression coverage for same-session workflow context, role filtering, recency, oversized response truncation and provider hostname rendering.
+
+## [5.64.0] - 2026-07-13 — Secure unified agent and operations foundation
+
+### Added
+- Unified `AgentCore`, `ToolExecutionGateway`, single tool manifest, mandatory JSON Schema validation and standardized tool results.
+- Conflict-safe structural patching, atomic rollback, coding `/undo`, `/retry`, `/compact`, local code intelligence and LSP JSON-RPC client.
+- Provenance/confidence/TTL memory records, conflict resolution, approved skill promotion, versioning and rollback.
+- Shared channel SDK with pairing-by-default, persistent idempotent delivery queue, retries, dead letters, reconnect supervision, health and metrics.
+- macOS Keychain, Windows DPAPI, Linux Secret Service and AES-256-GCM encrypted secret fallback.
+- Encrypted multi-device sync primitives with authenticated envelopes and vector-clock conflict detection.
+- Isolated sub-agent worktrees, test-failure analysis/repair loop, startup benchmark and TR/EN catalog snapshots.
+
+### Security
+- Fixed bulk-write guardrail bypasses, protected sensitive paths across execution origins and enabled guardrail hard-stop.
+- Removed high-risk shell string interpolation from iMessage delivery, backups, plugin installation/cloning and AI-generated Git commits.
+- Added channel pairing gates, non-interactive fail-closed permission handling and process-listener cleanup.
+
+### Tests
+- 73 test files, 711 passing tests (3 skipped), zero high-severity audit findings, CLI smoke and package checks.
+- Windows `--version` median reduced from ~366 ms to ~75 ms, meeting the <100 ms target.
+- Real Windows/macOS/Linux GitHub Actions matrix with lint, tests, smoke, audit and package verification.
+
 ## [5.63.0] - 2026-07-12 — Complete English REPL localization
 
 ### Changed

package/README.md

@@ -12,7 +12,7 @@
 
 A terminal-native AI agent CLI — chat, write code, automate workflows, and connect **Telegram / Discord / Slack / WhatsApp / iMessage** and more.
 
-**A Claude Code & OpenClaw alternative** · Multi-agent orchestration · Cross-session memory · Dangerous-command approval · 12 providers, 200+ models · 57+ tools · 10 messaging channels.
+**A Claude Code & OpenClaw alternative** · Multi-agent orchestration · Cross-session memory · Token-budgeted context · Dangerous-command approval · 12 providers, 200+ models · 106 tools · 10 messaging channels.
 
 ```
 ███╗   ██╗ █████╗ ████████╗██╗   ██╗██████╗ ███████╗ ██████╗  ██████╗
@@ -51,6 +51,8 @@ natureco code
 
 | Version | Highlights |
 |---------|-----------|
+| **v5.64.1** | **Reliable follow-ups + lower token cost:** `natureco code` now preserves same-session context across workflow calls and caps repeated history by token budget (1,024 / 2,048 / 8,192). Provider labels are rendered correctly. |
+| **v5.64.0** | **Unified secure agent foundation:** one execution gateway, hard-stop guardrails, schema-validated tools, rollback/checkpoints, sourced memory, resilient channel delivery, OS keychains and encrypted sync. |
 | **v5.43.0** | **Security:** 9 vulnerabilities fixed in a 3-round audit (RCE chain, admin-rpc auth, cron persistence, channel access control). See [`SECURITY_AUDIT_SUMMARY.md`](SECURITY_AUDIT_SUMMARY.md). |
 | **v5.42.0** | **Token optimization** — prompts trimmed by ~76% (skill index made compact; big cost savings on multi-step tasks). |
 | **v5.41.0** | **Multi-agent orchestration** — the agent can spawn focused sub-agents (`sub_agent`) and produce step-by-step plans (`plan`) before acting. |
@@ -102,11 +104,12 @@ Two-tier policy (`deny` / `allowlist` / `full`) applies to **every** shell path
 ## ✨ Features
 
 ### 🤖 AI & Chat
-- **57+ tools** — file ops, web search, image generation, code execution, memory, and more
+- **106 tools** — file ops, web search, image generation, code execution, memory, automation, channels, and more
 - **Interactive REPL** — read_file, edit_file, bash, multi-turn conversation
 - **Slash commands** — `/memory`, `/help`, `/skills`, `/model`, `/clear`
 - **Streaming output** with live tool-call visibility and a thinking indicator
 - **Persistent memory** — fact-based, cross-session
+- **Token-budgeted context** — recent relevant turns are retained without repeatedly sending oversized code and tool output
 
 ### 💻 Coding Agent (Claude Code alternative)
 - **Read / Write / Edit** multi-file operations
@@ -114,6 +117,20 @@ Two-tier policy (`deny` / `allowlist` / `full`) applies to **every** shell path
 - **Multi-agent orchestration** — spawn focused sub-agents and plan before acting
 - **Skills** — progressive-disclosure expertise loaded on demand via `skill_view`
 - **Verify loop** — the agent runs and tests the code it writes
+- **Reliable follow-ups** — references such as “the game you just created” retain the current coding-session context
+- **Context profiles** — Efficient (1,024), Balanced (2,048), and Quality (8,192) workflow-history token budgets
+
+### ⚡ Token Economy
+
+NatureCo uses progressive disclosure and bounded context instead of sending every skill, tool result, and old response on every request:
+
+- Skills are discovered with `skill_find` and loaded only when needed with `skill_view`.
+- System, internal tool, and empty messages are excluded from workflow history.
+- Recent user/assistant turns are kept within the selected token profile.
+- Oversized generated code is truncated in later prompts while the file path and conversational intent remain available.
+- `natureco ask` stays tool-free by default; use `--tools` only when an action is required.
+
+For a 32,000-character previous response, the Balanced profile bounds repeated history to about 2,048 tokens instead of roughly 8,000—a reduction of approximately 74% for that repeated context.
 
 ### 📡 10 Messaging Channels
 
@@ -161,7 +178,7 @@ Two-tier policy (`deny` / `allowlist` / `full`) applies to **every** shell path
 
 | Command | Description |
 |---------|-------------|
-| `natureco chat` | Interactive REPL chat (57+ tools active) |
+| `natureco chat` | Interactive REPL chat (106 tools available) |
 | `natureco chat --resume` | Resume the previous session |
 | `natureco code` | Coding agent (write apps/scripts) |
 | `natureco code <file>` | Coding agent on a specific file |

package/bin/natureco.js

@@ -1,6 +1,13 @@
-#!/usr/bin/env node
-
-const { Command } = require('commander');
+#!/usr/bin/env node
+
+// Fast path: version checks are used by shells, installers and health probes.
+// Avoid loading Commander and ~100 command modules for a constant-time answer.
+if (process.argv.length === 3 && (process.argv[2] === '--version' || process.argv[2] === '-V')) {
+  process.stdout.write(require('../package.json').version + '\n');
+  process.exit(0);
+}
+
+const { Command } = require('commander');
 const chalk = require('chalk');
 const packageJson = require('../package.json');
 // Küresel çökme/EPIPE yakalayıcıları — ham Node stack yerine düzgün mesaj + audit log

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "natureco-cli",
-  "version": "5.63.0",
+  "version": "5.64.1",
   "description": "Terminal-native AI agent CLI with bilingual TR/EN UI, multi-agent orchestration, persistent memory, secure tools and messaging integrations.",
   "bin": {
     "natureco": "bin/natureco.js"
@@ -22,10 +22,12 @@
     "test:unit": "vitest run",
     "test:watch": "vitest",
     "test:coverage": "vitest run --coverage",
+    "perf:startup": "node scripts/benchmark-startup.js",
     "lint": "eslint src/ bin/ test/",
     "lint:fix": "eslint src/ bin/ test/ --fix",
     "lint:errors-only": "eslint src/ bin/ test/ --quiet",
     "smoke": "node --check bin/natureco.js && node bin/natureco.js help",
+    "smoke:context": "node scripts/e2e-context-smoke.js",
     "postinstall": "node scripts/postinstall.js",
     "prepublishOnly": "node --check bin/natureco.js && eslint src/ bin/ test/ --quiet && vitest run"
   },

package/scripts/benchmark-startup.js

@@ -0,0 +1,26 @@
+'use strict';
+
+const { spawnSync } = require('child_process');
+const path = require('path');
+
+const bin = path.join(__dirname, '..', 'bin', 'natureco.js');
+function measure(args, runs = 7) {
+  const samples = [];
+  for (let i = 0; i < runs; i++) {
+    const start = process.hrtime.bigint();
+    const result = spawnSync(process.execPath, [bin, ...args], { encoding: 'utf8', env: { ...process.env, NATURECO_NO_UPDATE_CHECK: '1', FORCE_COLOR: '0' } });
+    const durationMs = Number(process.hrtime.bigint() - start) / 1e6;
+    if (result.status !== 0) throw new Error(`${args.join(' ')} failed: ${result.stderr}`);
+    samples.push(durationMs);
+  }
+  samples.sort((a, b) => a - b);
+  return { min: samples[0], median: samples[Math.floor(samples.length / 2)], p95: samples[Math.min(samples.length - 1, Math.ceil(samples.length * 0.95) - 1)], samples };
+}
+
+const version = measure(['--version']);
+const report = { timestamp: new Date().toISOString(), node: process.version, platform: process.platform, version };
+console.log(JSON.stringify(report, null, 2));
+if (version.median >= 100) {
+  console.error(`--version median ${version.median.toFixed(1)}ms exceeds 100ms target`);
+  process.exitCode = 1;
+}

package/scripts/e2e-context-smoke.js

@@ -0,0 +1,33 @@
+'use strict';
+
+const workflow = require('../src/tools/workflow');
+
+async function main() {
+  const marker = `natureco-context-${Date.now()}.html`;
+  const firstTask = `Bu test oturumundaki dosya adi ${marker}. Arac calistirma; sadece dosya adini iceren kisa bir onay ver.`;
+  const first = await workflow.execute({ action: 'run', task: firstTask });
+  if (!first || first.success === false || !first.reply) {
+    throw new Error(`First turn failed: ${first?.error || 'empty reply'}`);
+  }
+
+  const history = [
+    { role: 'user', content: firstTask },
+    { role: 'assistant', content: String(first.reply) },
+  ];
+  const second = await workflow.execute({
+    action: 'run',
+    task: 'Az once konustugumuz dosyanin tam adi neydi? Yalnizca dosya adini soyle.',
+    conversationHistory: history,
+  });
+  const reply = String(second?.reply || '');
+  if (!second || second.success === false || !reply.includes(marker)) {
+    throw new Error(`Context recall failed. Expected ${marker}; received: ${reply.slice(0, 240)}`);
+  }
+
+  console.log(JSON.stringify({ ok: true, marker, reply: reply.trim() }));
+}
+
+main().catch(error => {
+  console.error(error.message);
+  process.exitCode = 1;
+});

package/src/commands/backup.js

@@ -4,7 +4,7 @@ const { getLang: _gl } = require('../utils/i18n');
 const L = (tr, en) => (_gl() === 'en' ? en : tr);
 const path = require('path');
 const os = require('os');
-const { execSync } = require('child_process');
+const { execFileSync } = require('child_process');
 
 const NATURECO_DIR = path.join(os.homedir(), '.natureco');
 
@@ -32,7 +32,7 @@ function createBackup() {
 
   if (fs.existsSync(NATURECO_DIR)) {
     try {
-      execSync(`tar -czf "${backupFile}" -C "${path.dirname(NATURECO_DIR)}" "${path.basename(NATURECO_DIR)}"`, {
+      execFileSync('tar', ['-czf', backupFile, '-C', path.dirname(NATURECO_DIR), path.basename(NATURECO_DIR)], {
         stdio: 'pipe',
         encoding: 'utf8',
         timeout: 30000
@@ -97,7 +97,7 @@ function restoreBackup(file) {
 
   console.log(chalk.yellow(`\n  ⚠️  Restoring ${file}...\n`));
   try {
-    execSync(`tar -xzf "${backupFile}" -C "${os.homedir()}"`, { stdio: 'pipe', timeout: 15000 });
+    execFileSync('tar', ['-xzf', backupFile, '-C', os.homedir()], { stdio: 'pipe', timeout: 15000 });
     console.log(chalk.green('  ✅ Restore complete\n'));
   } catch {
     console.log(chalk.red('  ❌ Restore failed\n'));

package/src/commands/code.js

@@ -2,7 +2,7 @@ const path = require('path');
 const os = require('os');
 const fs = require('fs');
 const readline = require('readline');
-const { execSync } = require('child_process');
+const { execSync, execFileSync } = require('child_process');
 const inquirer = require('../utils/inquirer-wrapper');
 const TB = require('../utils/token-budget');
 const chalk = require('chalk');
@@ -14,7 +14,11 @@ const { getMemoryPrompt, loadMemory } = require('../utils/memory');
 const { getAgentsPrompt } = require('../utils/agents');
 const { createSession, addMessageToSession } = require('../utils/sessions');
 const { addToHistory } = require('../utils/history');
-const { getToolDefinitions, executeTool } = require('../utils/tool-runner');
+const { getToolDefinitions, executeTool } = require('../utils/tool-runner');
+const { AgentCore } = require('../utils/agent-core');
+const { CodingSession } = require('../utils/coding-session');
+
+const agentCore = new AgentCore({ maxIterations: 10 });
 
 let rl = null;
 
@@ -224,12 +228,17 @@ async function streamMessage(providerConfig, messages, tools) {
 // ── Tool execution ────────────────────────────────────────────────────────────
 const DANGEROUS = [/\brm\b/, /\brmdir\b/, /\bdelete\b/i, /\bdrop\b/i, /\btruncate\b/i];
 
-async function runToolCall(toolCall, stats, dryRun = false) {
+async function runToolCall(toolCall, stats, dryRun = false, codingSession = null) {
+  const guard = agentCore.assess({ name: toolCall.name, input: toolCall.input });
+  if (guard.blocked) return { success: false, error: guard.reason || `blocked_by_guardrails: ${toolCall.name}` };
   const inputPreview = JSON.stringify(toolCall.input).slice(0, 60);
 
-  const needsConfirm =
+  const needsConfirm =
     toolCall.name === 'write_file' ||
-    (toolCall.name === 'bash' && DANGEROUS.some(re => re.test(toolCall.input.command || '')));
+    (toolCall.name === 'bash' && DANGEROUS.some(re => re.test(toolCall.input.command || '')));
+
+  const risk = codingSession?.riskSummary(toolCall);
+  if (risk && risk.level !== 'low') console.log(chalk.gray(`  Risk: ${risk.level} (${risk.risks.join(', ')})`));
 
   // Dry-run: write_file'ı engelle, diff göster
   if (dryRun && toolCall.name === 'write_file') {
@@ -261,8 +270,12 @@ async function runToolCall(toolCall, stats, dryRun = false) {
     }
   }
 
-  const spinner = startSpinner(`${toolCall.name}  ${inputPreview}`);
-  const result = await executeTool(toolCall.name, toolCall.input);
+  const spinner = startSpinner(`${toolCall.name}  ${inputPreview}`);
+  if (!dryRun && codingSession && (toolCall.name === 'write_file' || toolCall.name === 'edit_file')) {
+    codingSession.capture(toolCall.input.path);
+  }
+  const result = await executeTool(toolCall.name, toolCall.input);
+  agentCore.record({ name: toolCall.name, input: toolCall.input }, result);
   stopSpinner(spinner, toolCall.name, result.success !== false);
 
   if (result.success !== false) {
@@ -310,7 +323,8 @@ async function runTests(projectIndex, conversationMessages, tools, providerConfi
 }
 
 // ── Main ──────────────────────────────────────────────────────────────────────
-async function code(targetFile, options = {}) {
+async function code(targetFile, options = {}) {
+  const codingSession = new CodingSession();
   const workDir = options.dir || process.cwd();
   const apiKey = getApiKey();
   const config = getConfig();
@@ -507,15 +521,35 @@ ${indexPrompt}`);
             ['/run <cmd>',L('Komutu çalıştır, hata varsa düzelt', 'Run command, fix errors if any')],
             ['/test',    L('Testleri çalıştır, hata varsa düzelt', 'Run tests, fix errors if any')],
             ['/git',     L('Git durumu ve son commitler', 'Git status and recent commits')],
-            ['/commit',  L('Staged değişiklikleri AI ile commit et', 'Commit staged changes with AI')],
+            ['/commit',  L('Staged değişiklikleri AI ile commit et', 'Commit staged changes with AI')],
+            ['/undo',    L('Son dosya değişikliğini geri al', 'Undo the last file change')],
+            ['/retry',   L('Son isteği yeniden çalıştır', 'Retry the last request')],
+            ['/compact', L('Konuşma bağlamını sıkıştır', 'Compact conversation context')],
             ['/help',    L('Bu yardım', 'This help')],
           ].forEach(([c, d]) => console.log('  ' + chalk.cyan(c.padEnd(12)) + chalk.gray(d)));
           console.log(chalk.gray('  Ctrl+C'.padEnd(14) + L('Çıkış', 'Exit')));
           console.log();
           return;
-        case 'clear':
+        case 'clear':
           console.clear();
-          return;
+          return;
+        case 'undo': {
+          const undone = codingSession.undo();
+          console.log(undone.ok ? chalk.green(`  ✓ ${undone.path}`) : chalk.yellow(`  ${undone.error}`));
+          return;
+        }
+        case 'compact': {
+          const compacted = codingSession.compact(conversationMessages);
+          conversationMessages.splice(0, conversationMessages.length, ...compacted.messages);
+          console.log(chalk.green(`  ✓ Context compacted: ${compacted.before} → ${compacted.after}`));
+          return;
+        }
+        case 'retry': {
+          const previous = codingSession.retryMessage();
+          if (!previous) console.log(chalk.yellow(L('  Tekrarlanacak istek yok.', '  No request to retry.')));
+          else await handleMessage(previous);
+          return;
+        }
         case 'summary':
         case 'done': {
           const sum = await generateSummary(conversationMessages, providerConfig);
@@ -591,7 +625,7 @@ ${indexPrompt}`);
             console.log(chalk.cyan(`\n  ${L('Önerilen', 'Suggested')}: ${chalk.white(commitMsg)}\n`));
             const ok = await confirmAction(L('Commit edilsin mi?', 'Commit?'));
             if (ok) {
-              execSync(`git commit -m "${commitMsg.replace(/"/g, '\\"')}"`, { cwd: workDir, stdio: 'pipe' });
+              execFileSync('git', ['commit', '-m', commitMsg], { cwd: workDir, stdio: 'pipe' });
               console.log(chalk.green(L('  ✓ Commit yapıldı!\n', '  ✓ Committed!\n')));
             } else {
               console.log(chalk.gray(L('  İptal edildi.\n', '  Cancelled.\n')));
@@ -607,7 +641,8 @@ ${indexPrompt}`);
       }
     }
 
-    stats.messageCount++;
+    stats.messageCount++;
+    codingSession.rememberUserMessage(userMessage);
     console.log(chalk.white('You  ') + userMessage);
     conversationMessages.push({ role: 'user', content: userMessage });
 
@@ -655,7 +690,7 @@ ${indexPrompt}`);
 
       for (let ti = 0; ti < streamResult.toolCalls.length; ti++) {
         const toolCall = streamResult.toolCalls[ti];
-        const result = await runToolCall(toolCall, stats, options.dryRun);
+        const result = await runToolCall(toolCall, stats, options.dryRun, codingSession);
         const resultStr = result.success !== false
           ? (result.output || JSON.stringify(result))
           : `${L('Hata', 'Error')}: ${result.error}`;

package/src/commands/code_v5.js

@@ -33,6 +33,11 @@ const { getFallbackChain } = require("../utils/fallback-chain");
 const { getTaskManager } = require("../utils/tasks");
 const { buildTiers, assemble, discoverProjectRules } = require("../utils/system-prompt");
 const { buildSkillIndex } = require("../utils/skill-index");
+const { AgentCore } = require("../utils/agent-core");
+const { prepareConversationHistory } = require("../utils/conversation-context");
+const tokenBudget = require("../utils/token-budget");
+
+const agentCore = new AgentCore({ maxIterations: 10 });
 
 const IS_MAC = os.platform() === "darwin";
 const MAX_ITERATIONS = 10;
@@ -519,7 +524,14 @@ async function codeV5(targetPath) {
 
       // Workflow step — run before every request
       process.stdout.write(tui.styled('\r  🔧 workflow...  ', { color: tui.PALETTE.muted }));
-      const wfResult = await executeTool('workflow', { action: 'run', task: input }, toolDefs);
+      const wfResult = await executeTool('workflow', {
+        action: 'run',
+        task: input,
+        conversationHistory: prepareConversationHistory(messages, {
+          maxMessages: tokenBudget.load().conversationInContext,
+          maxTokens: tokenBudget.load().workflowHistoryMaxTokens,
+        }),
+      }, toolDefs);
       const wf = wfResult?.result || {};
       if (wf.success !== false) {
         const loaded = wf.skillsLoaded && wf.skillsLoaded.length > 0 ? ` [skill: ${wf.skillsLoaded.join(', ')}]` : '';
@@ -624,9 +636,16 @@ async function codeV5(targetPath) {
 const PARALLEL_SAFE_TOOLS = new Set(['read_file', 'file_search', 'grep_search', 'web_search', 'web_readability', 'duckduckgo_search', 'exa_search', 'searxng_search', 'firecrawl', 'memory_search', 'memory']);
 
 async function processToolCalls(reply, config, toolDefs, messages, onToolResult) {
+  agentCore.startIteration();
+  const coreBlocked = new Map();
   for (const tc of reply.tool_calls) {
     let args = {};
     try { args = JSON.parse(tc.function.arguments || "{}"); } catch {}
+    const guard = agentCore.assess({ name: tc.function.name, input: args });
+    if (guard.blocked) {
+      coreBlocked.set(tc.id, guard.reason || "blocked_by_guardrails");
+      continue;
+    }
     const risk = assessRisk(tc.function.name, args);
 
     // Built-in risk assessment
@@ -687,7 +706,11 @@ async function processToolCalls(reply, config, toolDefs, messages, onToolResult)
 
   messages.push({ role: "assistant", content: reply.content || null, tool_calls: reply.tool_calls });
 
-  const parsed = reply.tool_calls.map(tc => {
+  for (const [id, reason] of coreBlocked) {
+    messages.push({ role: "tool", tool_call_id: id, content: "ERROR: " + reason });
+  }
+
+  const parsed = reply.tool_calls.filter(tc => !coreBlocked.has(tc.id)).map(tc => {
     let args = {};
     try { args = JSON.parse(tc.function.arguments || "{}"); } catch {}
     return { name: tc.function.name, args, id: tc.id };