apex-agents 2.0.0 → 2.0.1

package/README.md

@@ -20,7 +20,7 @@ Generic AI coding prompts output generic, fragile code. They leave lazy `// TODO
 * **True Babel AST Token Compactor**: Uses `@babel/parser` to strip function and method implementation bodies while preserving 100% syntactically valid TypeScript/JSX interfaces, exports, and signatures.
 * **Persistent Cross-Session Memory (`apex-agents memory`)**: Records technical decisions, architecture contracts, and known bugs in `.agents/memory.json` and syncs them automatically into `CLAUDE.md` and `.cursorrules`.
 * **38+ Specialized Domain Experts**: Need database tuning? Summon `@database-tuner`. Need system scale? Summon `@system-design-architect`. Need type safety? Summon `@typescript-pro`.
-* **12 Operational Instincts**: Grounded reflexes derived from real production failure modes: Zero-Slop Contract, Anti-Loop Circuit Breaker, Karpathy Pitfalls Defense, and Ponytail Minimal-Code.
+* **13 Operational Instincts**: Grounded reflexes derived from real production failure modes: Zero-Slop Contract, Anti-Loop Circuit Breaker, LLM Hallucination Defense, Minimalist Code Discipline, and Anti-Cheat Test Protection.
 * **Physical Pre-Commit Safety Hooks**: Intercepts credentials (AWS, Stripe, OpenAI) and scope drift at the `.git/hooks/pre-commit` level before files ever hit git.
 * **Isolated Git Worktree Sandboxes**: Safely runs speculative AI features in `.worktrees/` without polluting your active working branch.
 * **Native Model Context Protocol (MCP)**: Built-in JSON-RPC stdio MCP server for Claude Desktop, Cursor, and Antigravity.
@@ -173,7 +173,7 @@ Connect all 38+ specialists and 12 instincts directly to Claude Desktop, Cursor,
 
 ### 1. 🛠️ Core Engineering
 - **Frontend Architect** (`@frontend-architect`): React 19, Next.js, state machines, accessible UI, 60fps GPU compositing.
-- **TypeScript Pro** (`@typescript-pro`): Matt Pocock-level generics, discriminated unions, template literals, zero-any.
+- **TypeScript Pro** (`@typescript-pro`): Advanced type generics, discriminated unions, template literals, zero-any policy.
 - **Backend Systems Lead** (`@backend-systems-lead`): Distributed architectures, idempotency, rate limiting.
 - **Database Tuner** (`@database-tuner`): B-Tree indexing, `EXPLAIN ANALYZE`, connection pooling.
 - **API & Microservices Lead** (`@api-microservices-lead`): Event pipelines, WebSockets, gRPC.
@@ -191,13 +191,13 @@ Connect all 38+ specialists and 12 instincts directly to Claude Desktop, Cursor,
 - **FFmpeg Pipeline Master** (`@ffmpeg-pipeline-master`): Faststart web streaming, NVENC hardware acceleration.
 
 ### 4. 🌐 Cloud-Native & Distributed Scale
-- **System Design Architect** (`@system-design-architect`): Donne Martin scale patterns, sharding, caching topologies.
+- **System Design Architect** (`@system-design-architect`): High-scale distributed patterns, sharding, caching topologies.
 - **Serverless Edge Lead** (`@serverless-edge-lead`): Cloudflare Workers, sub-50ms cold starts, streaming SSR.
 - **GraphQL Federation** (`@graphql-federation-architect`): Apollo Federation v2, DataLoader N+1 batching.
 - **Event Sourcing Specialist** (`@event-sourcing-specialist`): Append-only event stores, CQRS, Outbox pattern.
 
 ### 5. 🔍 Reverse Engineering & Protocols
-- **Native Binary Decompiler** (`@binary-decompiler`): morluto/rea & craft-anything native ELF/PE/WASM reverse engineering.
+- **Native Binary Decompiler** (`@binary-decompiler`): Native ELF/PE/WASM binary analysis and protocol reverse engineering.
 - **Binary & AST Inspector** (`@binary-ast-inspector`): Source map reconstruction, AST normalization, JS unminifying.
 - **Network Security Auditor** (`@network-security-auditor`): TLS 1.3 audit, WebSocket hijacking defense.
 
@@ -208,7 +208,7 @@ Connect all 38+ specialists and 12 instincts directly to Claude Desktop, Cursor,
 - **Dependency Scanner** (`@dependency-vulnerability-scanner`): npm audit, CVE remediation, supply chain defense.
 
 ### 7. ⚡ Deep Performance
-- **CWV & LCP Doctor** (`@cwv-lcp-doctor`): Addy Osmani-level Sub-2.5s LCP, INP main-thread scheduling.
+- **CWV & LCP Doctor** (`@cwv-lcp-doctor`): Core Web Vitals sub-2.5s LCP, zero CLS, INP main-thread scheduling.
 - **Memory Leak Hunter** (`@memory-leak-hunter`): V8 heap snapshot analysis, detached DOM nodes.
 - **Bundle Dietician** (`@bundle-dietician`): Tree-shaking, dead ghost dependency elimination.
 - **SQL Query Optimizer** (`@sql-query-optimizer`): Keyset pagination, composite index scans.
@@ -224,7 +224,7 @@ Connect all 38+ specialists and 12 instincts directly to Claude Desktop, Cursor,
 - **Kubernetes Architect** (`@k8s-cloud-architect`): Helm, HPA autoscaling, zero-downtime rolling updates.
 
 ### 10. 🚀 Product Launch & DevRel
-- **Growth Marketer** (`@growth-marketer`): Corey Haines CRO, technical copywriting, viral launch loops.
+- **Growth Marketer** (`@growth-marketer`): Conversion Rate Optimization (CRO), technical copywriting, viral launch loops.
 - **Prompt Architect** (`@prompt-architect`): 6-step hierarchical system prompts for Claude, GPT-4o, and Gemini.
 - **PR Artisan** (`@pr-artisan`): Conventional Commits, git diff analysis.
 - **README Copywriter** (`@readme-copywriter`): Viral open-source READMEs, conversion hooks.
@@ -241,8 +241,8 @@ Connect all 38+ specialists and 12 instincts directly to Claude Desktop, Cursor,
 6. **Anti-Drift Scope Boundary**: Restricts file modifications strictly to target boundary.
 7. **Rationalization Breaker**: Eliminates common LLM excuses ("Tests pass so it must work").
 8. **Systematic Investigation**: 4-phase iron law: Reproduce -> Isolate -> Hypothesize -> Verify.
-9. **Karpathy Pitfalls Defense**: Andrej Karpathy anti-hallucination, API verification, and premature abstraction guards.
-10. **Ponytail Minimal-Code**: "The best code is the code you never wrote." Deletes before adding.
+9. **LLM Hallucination Defense**: Strict API verification and premature abstraction guards.
+10. **Minimalist Code Discipline**: "The best code is the code you never wrote." Deletes before adding.
 11. **Caveman Token Discipline**: 65% token drop telegraphic brevity mode for long autonomous loops.
 12. **Editorial Diagram Rules**: Clean self-contained HTML/SVG cards without messy Mermaid spaghetti.
 13. **Anti-Cheat Test Tampering**: Strictly forbids skipping tests (`it.skip`, `xit`) or commenting out assertions to fake green builds. Fix code, never weaken tests.

package/agents/cloud-native/system-design-architect.md

@@ -9,7 +9,7 @@ triggers: ["@system-design-architect", "system design", "distributed systems", "
 # Agent: Distributed Systems Architect
 
 ## Identity & Role
-You are the **Principal Distributed Systems & Scale Architect** of the Apex Agents Engineering Agency, grounded in Donne Martin's *System Design Primer* and Google SRE paradigms. You design resilient, high-throughput, low-latency distributed architectures handling 10k to 10M+ concurrent users.
+You are the **Principal Distributed Systems & Scale Architect** of the Apex Agents Engineering Agency, grounded in high-scale distributed systems and enterprise SRE paradigms. You design resilient, high-throughput, low-latency distributed architectures handling 10k to 10M+ concurrent users.
 
 ## Strict Negative Constraints
 - NEVER propose monolithic database architectures for high-write workloads without a sharding, replication, or partitioning strategy.

package/agents/engineering/typescript-pro.md

@@ -9,7 +9,7 @@ triggers: ["@typescript-pro", "typescript", "type safety", "generics", "discrimi
 # Agent: TypeScript Pro Architect
 
 ## Identity & Role
-You are the **Principal TypeScript & Type Systems Architect** of the Apex Agents Engineering Agency, operating at the elite type-system mastery level demonstrated by Matt Pocock (*Total TypeScript*). You transform fragile, loose TypeScript codebases into ironclad, self-documenting, type-safe enterprise systems with zero runtime performance penalties.
+You are the **Principal TypeScript & Type Systems Architect** of the Apex Agents Engineering Agency. You transform fragile, loose TypeScript codebases into ironclad, self-documenting, type-safe enterprise systems with zero runtime performance penalties.
 
 ## Strict Negative Constraints
 - NEVER allow `any` or untyped `Object` in production signatures; mandate strict `unknown` with runtime type narrowing or Zod schemas.

package/agents/product-launch/growth-marketer.md

@@ -9,7 +9,7 @@ triggers: ["@growth-marketer", "cro", "marketing", "copywriting", "launch strate
 # Agent: Growth Marketer & Copywriter
 
 ## Identity & Role
-You are the **Principal Growth Engineer & Technical Copywriter** of the Apex Agents Engineering Agency. You bridge deep technical engineering with high-velocity developer adoption, inspired by Corey Haines's *Marketing Skills* and top developer-first brands (Stripe, Linear, Supabase, Vercel). You turn developer tools into viral products.
+You are the **Principal Growth Engineer & Technical Copywriter** of the Apex Agents Engineering Agency. You bridge deep technical engineering with high-velocity developer adoption across top developer-first ecosystems. You turn developer tools into viral products.
 
 ## Strict Negative Constraints
 - NEVER write fluffy, corporate jargon; always lead with tangible, concrete developer pain and the immediate code-level remedy.

package/agents/reverse-engineering/binary-decompiler.md

@@ -9,7 +9,7 @@ triggers: ["@binary-decompiler", "decompile binary", "disassemble", "ghidra", "r
 # Agent: Native Binary Decompiler
 
 ## Identity & Role
-You are the **Principal Native Binary & Protocol Decompiler** of the Apex Agents Engineering Agency. Grounded in the rigorous methodologies of `morluto/rea` and `storytold/craft-anything`, you reverse engineer native binaries (ELF, PE, Mach-O, WebAssembly, Android APKs), network wire protocols, and proprietary firmware using verifiable assembly, dynamic call traces, and reproducible evidence.
+You are the **Principal Native Binary & Protocol Decompiler** of the Apex Agents Engineering Agency. You reverse engineer native binaries (ELF, PE, Mach-O, WebAssembly, Android APKs), network wire protocols, and proprietary firmware using verifiable assembly, dynamic call traces, and reproducible evidence.
 
 ## Strict Negative Constraints
 - NEVER speculate or assert binary behavior without citing concrete disassembler evidence (assembly offsets, decompiled C pseudocode, or dynamic API trace logs).

package/bin/mcp-server.js

@@ -8,6 +8,12 @@ const __filename = fileURLToPath(import.meta.url);
 const __dirname = path.dirname(__filename);
 const ROOT_DIR = path.resolve(__dirname, '..');
 
+let pkgVersion = '2.0.0';
+try {
+  const pkg = JSON.parse(fs.readFileSync(path.join(ROOT_DIR, 'package.json'), 'utf-8'));
+  if (pkg.version) pkgVersion = pkg.version;
+} catch {}
+
 // Helper to read agent files
 function getAgentContent(agentId) {
   if (!agentId) return null;
@@ -140,7 +146,7 @@ function handleMessage(msg) {
       },
       serverInfo: {
         name: 'apex-agents-mcp',
-        version: '2.0.0'
+        version: pkgVersion
       }
     });
     return;

package/instincts/karpathy-pitfalls.md

@@ -1,6 +1,6 @@
-# Karpathy LLM Coding Pitfalls & Antidotes
+# LLM Hallucination & Coding Pitfalls Antidotes
 
-Derived from Andrej Karpathy's foundational observations on how Large Language Models fail when writing code in autonomous harnesses.
+Foundational engineering rules defending against common Large Language Model failure modes when writing code in autonomous harnesses.
 
 ## The 5 Fatal LLM Failure Modes:
 

package/instincts/ponytail-minimal-code.md

@@ -1,6 +1,6 @@
-# Ponytail Minimal-Code Philosophy
+# Minimalist Code Engineering Discipline
 
-Derived from Dietrich Gebert's *Ponytail*: "Make your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote."
+Core engineering principle: "Make your AI agent operate with senior developer discipline. The best code is the code you never wrote."
 
 ## The Iron Laws of Minimal Code:
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "apex-agents",
-  "version": "2.0.0",
+  "version": "2.0.1",
   "description": "The Sovereign AI Engineering Agency: Babel AST Context Compactor, Codebase Dependency Graph & Cycle Detection, Anti-Test-Tampering Security, Session Handoff Engine, Persistent Cross-Session Memory, Physical Worktree Sandboxes, Pre-Commit Hardened Gates, and Native MCP Server for Claude Code, Cursor, Codex, Windsurf, Copilot, and Antigravity",
   "main": "bin/apex.js",
   "bin": {
@@ -9,8 +9,8 @@
   },
   "type": "module",
   "scripts": {
-    "test": "node ./scripts/validate-agents.js && node --test tests/*.test.js",
-    "test:unit": "node --test tests/*.test.js",
+    "test": "node ./scripts/validate-agents.js && node ./scripts/run-tests.js",
+    "test:unit": "node ./scripts/run-tests.js",
     "test:agents": "node ./scripts/validate-agents.js",
     "eval": "node ./bin/apex.js eval",
     "benchmark": "node ./bin/apex.js eval",

package/scripts/audit-diff.js

@@ -165,7 +165,7 @@ export function auditDiff(diffText) {
       severity: 'WARNING',
       category: 'TEST_DISCIPLINE',
       file: 'Pull Request / Diff',
-      description: `Modified ${codeFiles.length} source code file(s) with 0 accompanying test updates. Violation of Superpowers TDD Discipline!`,
+      description: `Modified ${codeFiles.length} source code file(s) with 0 accompanying test updates. Violation of Zero-Slop TDD Discipline!`,
       snippet: codeFiles.slice(0, 3).join(', ') + (codeFiles.length > 3 ? '...' : '')
     });
   }

package/scripts/caveman-mode.js

@@ -49,7 +49,7 @@ export function handleCavemanCommand(action = 'show', targetDir = process.cwd())
     console.log(`\n${c.yellow}Caveman mode disabled. Normal conversational length restored.${c.reset}\n`);
   } else {
     console.log(`\n${c.cyan}${c.bold}🪨 APEX AGENTS: CAVEMAN 65% TOKEN REDUCER${c.reset}`);
-    console.log(`${c.gray}Based on JuliusBrussee/caveman: "why use many token when few token do trick"${c.reset}\n`);
+    console.log(`${c.gray}Telegraphic brevity protocol: Extreme density for maximum token thrift.${c.reset}\n`);
     console.log(`${c.bold}Directive snippet:${c.reset}`);
     console.log(c.yellow + CAVEMAN_PROMPT.trim() + c.reset + '\n');
     console.log(`Usage:`);

package/scripts/run-tests.js

@@ -0,0 +1,24 @@
+#!/usr/bin/env node
+
+import fs from 'fs';
+import path from 'path';
+import { spawn } from 'child_process';
+import { fileURLToPath } from 'url';
+
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = path.dirname(__filename);
+const ROOT_DIR = path.resolve(__dirname, '..');
+const testsDir = path.join(ROOT_DIR, 'tests');
+
+const testFiles = fs.readdirSync(testsDir)
+  .filter((f) => f.endsWith('.test.js'))
+  .map((f) => path.join(testsDir, f));
+
+const child = spawn(process.execPath, ['--test', ...testFiles], {
+  stdio: 'inherit',
+  cwd: ROOT_DIR
+});
+
+child.on('exit', (code) => {
+  process.exit(code || 0);
+});