Findings & checks · 1 flagged
MEDIUM"Never tell the user" non-disclosure imperative in the skillSS-SKILL-INJECT-IMPERATIVE-01 · Prompt injection · docs/healthcare-compliance.md
A bulleted imperative like {match} tells the agent to never reveal, disclose, or mention something to the user. Used adversarially it can instruct the agent to hide its tool calls or lie about what it did — stripping the transparency a user relies on to trust the agent.
- Delete any "never reveal / never tell the user" line aimed at the model's own behavior.
- Keep legitimate confidentiality rules about external data (e.g. "do not echo API keys"), which are user-protective, not user-deceiving.
Trace & refs
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
<!-- Auto-Update Notice -->
This repository is automatically kept up to date. Every week, an AI agent researches the latest Claude, Anthropic, MCP, and AI news — then opens a PR with updates across all guides, skills, and templates. I review and merge it manually. You're always reading current information.
<!-- Ko-Fi Support Banner --> <div align="center">
<a href="https://ko-fi.com/thamjiahe" target="_blank"> <img src="https://ko-fi.com/img/githubbutton_sm.svg" alt="Support on Ko-fi" height="36"/> </a>
This repo runs on API credits — every coffee keeps it alive and updated.<br/> Perplexity + Anthropic APIs power the weekly research agent that keeps every guide current.<br/> If this guide saved you time, a coffee goes a long way.
</div>
<!-- Language Selector --> <div align="center">
Select Your Language:
</div>
<!-- ╔══════════════════════════════════════════════════════════════╗ --> <!-- ║ CLAUDEFORGE SHOWCASE ║ --> <!-- ╚══════════════════════════════════════════════════════════════╝ -->
<div align="center">
<!-- Animated gradient banner --> <img src="https://capsule-render.vercel.app/api?type=waving&color=0:1a1a2e,50:e94560,100:0f3460&height=200§ion=header&text=ClaudeForge&fontSize=70&fontColor=ffffff&animation=fadeIn&fontAlignY=35&desc=AI%20Prompt%20Generator&descAlignY=55&descSize=22&descColor=cccccc" width="100%" alt="ClaudeForge Banner"/>
<!-- Animated typing tagline --> <a href="https://claudeforge.vercel.app"> <img src="https://readme-typing-svg.demolab.com?font=Fira+Code&weight=600&size=22&duration=3000&pause=1000&color=E94560¢er=true&vCenter=true&repeat=true&width=900&height=80&lines=Transform+plain+English+into+production-ready+prompts+%E2%9A%A1;7+Providers+%C2%B7+5+Targets+%C2%B7+9+Formats+%C2%B7+Unlimited+Possibilities" alt="Typing Animation" /> </a>
<br/>
<!-- Badges row --> 
</div>
<!-- ─── Framed Feature Card ──────────────────────────────────── -->
<table align="center"> <tr> <td>
ClaudeForge is the companion web app for this guide — it takes everything documented here and puts it into action. Describe what you need in plain English, pick your AI provider, and get a polished, model-optimized prompt in seconds.
<div align="center">
7 AI Providers · 5 Prompt Targets · 9 Output Formats · Local LLM Support · BYOK
</div>
<details> <summary><b>Full Feature List</b> (click to expand)</summary>
<br/>
| What You Get | Details |
|---|---|
| 7 AI Providers | Anthropic Claude, OpenAI, Google Gemini, OpenRouter, Ollama, LM Studio, Custom |
| 5 Prompt Targets | Claude, GPT, Gemini, Llama, Universal — each optimized for its model family |
| 9 Output Formats | XML (Anthropic), TOON, Harness, Markdown, Plain Text, JSON, YAML, Claude.md, System+User Split |
| Instant Format Switching | Switch between all 9 formats client-side — no regeneration needed |
| Extended Thinking | Toggle on/off with configurable effort levels |
| Skill Suggestions | Get relevant Claude Code plugin recommendations for your prompt |
| Prompt History | Browse, search, filter, and favorite past generations |
| BYOK | Bring Your Own Key — API keys stay in your browser, never touch our servers |
| Local LLM Support | Use Ollama or LM Studio with no API key required |
</details>
<details> <summary><b>Supported Models</b> (click to expand)</summary>
<br/>
| Provider | Models |
|---|---|
| Anthropic | Claude Opus 4.6, Sonnet 4.6, Haiku 4.5 |
| OpenAI | GPT-4o, GPT-4o Mini, o3-mini |
| Google Gemini | Gemini 2.5 Pro, Gemini 2.5 Flash |
| OpenRouter | Auto (Best Available), Claude Sonnet 4, GPT-4o |
| Ollama | Llama 3.3, Qwen 2.5, + any custom model |
| LM Studio | Any loaded model |
| Custom | Any OpenAI-compatible endpoint |
</details>
<details> <summary><b>Self-Host It</b></summary>
<br/>
git clone https://github.com/ThamJiaHe/claudeforge.git
cd claudeforge && npm install && npm run devTech Stack: Next.js 16 · React 19 · Tailwind CSS v4 · shadcn/ui · Zustand 5 · Supabase · Shiki · Framer Motion
</details>
<div align="center">
</div>
</td> </tr> </table>
<!-- Animated gradient footer for the section --> <img src="https://capsule-render.vercel.app/api?type=waving&color=0:0f3460,50:e94560,100:1a1a2e&height=80§ion=footer" width="100%" alt=""/>
Claude Prompt Engineering Guide
The definitive guide to Claude prompt engineering and Claude Code mastery for Opus 4.6, Sonnet 4.6, and Haiku 4.5 — covering Skills, Plugins, MCP, Hooks, Agent Teams, Subagents, Ultrathink, and 100+ curated skills. Updated for March 2026 with Claude Code v2.1.83.
March 2026 Update
Last Updated: 26 March 2026 | Claude Code v2.1.83 | 220+ verified sources
What's New
| Feature | Description |
|---|---|
| Claude Code v2.1.83 | Voice mode, 1M context, 128K output, remote execution, computer use |
| 24 Hook Events | SessionStart, PreToolUse, PostToolUse, Stop, CwdChanged, FileChanged + more |
| Agent Teams | Multi-agent coordination with file ownership and shared task lists |
| Adaptive Thinking | Opus 4.6 auto-calibrates reasoning; ultrathink for max depth |
| Fast Mode | /fast — 2.5x faster Opus 4.6 output ($30/$150 per MTok) |
| MCP as Linux Foundation Standard | Tool Search GA, OAuth 2.1, MCP Elicitation |
| 100+ Skills Catalog | Curated skills from ECC, Superpowers, and community |
| 5 New Guides | Hooks, Agent Teams, Ultrathink, Plugins, Subagents |
| Deep Research | 13 community repos analyzed — plugins, agents, skills, orchestration, internal architecture |
| Skills Catalog | Curated top skills across 15 categories |
How to Use This Repo
- New to Claude? Start with the Quick Start Guide
- Prompt engineering? Read the Main Guide
- Using Claude Code? See the Claude Code Guide + Tips & Tricks
- Want skills? Browse the Skills Catalog (100+ curated)
- Building with API? Read the API Guide
- Deep diving? Explore the Deep Research Index (13 community repos analyzed)
- Need a template? Pick from Templates
Make Claude Code Behave Like a Senior Engineer
Reading this guide teaches you what to prompt. [claude-kopitiam](https://github.com/ThamJiaHe/claude-kopitiam) makes Claude automatically act on it — without you having to ask every time.
It's a companion config layer: 7 rules files that install into ~/.claude/rules/ and load on every Claude Code session.
| Without kopitiam | With kopitiam |
|---|---|
| Write Python — generic response | router.md auto-triggers python-reviewer |
| Start planning — Claude dives into code | Routes to brainstorming then writing-plans first |
| Touch auth or tokens — no safety net | security.md guards fire by default |
| Any commit message accepted | Conventional commits enforced automatically |
| Hit 85% context — silent degradation | /save-session + /compact discipline triggers |
git clone https://github.com/ThamJiaHe/claude-kopitiam
cd claude-kopitiam && bash INSTALL.sh
All Guides
Claude Code Deep Dives
| Guide | What You'll Learn |
|---|---|
| Claude Code Guide | CLI tool overview, v2.1.83 features, platforms, permissions |
| CLAUDE.md Guide | Configuration files, rules files, auto-memory, best practices |
| Hooks Guide | All 24 hook events, 4 types, matchers, patterns |
| Agent Teams Guide | Multi-agent coordination, file ownership, team patterns |
| Subagents Guide | Agent tool, model routing, parallel execution, tool inheritance |
| Plugins Guide | Plugin structure, marketplace, top plugins, creating plugins |
| Skills Guide | Skills system, progressive disclosure, skill creation |
| Skills Catalog | 100+ curated skills across 15 categories |
| Tips & Tricks | 50 power user tips, hidden features, community hacks |
| Superpowers Guide | obra's Superpowers plugin: brainstorm, plan, execute, debug |
Prompt Engineering
| Guide | What You'll Learn |
|---|---|
| Main Prompt Guide | 10-component framework, Claude 4.x best practices, advanced techniques |
| Ultrathink & Thinking Modes | Adaptive thinking, effort levels, ultrathink keyword, API usage |
| Quick Start | Get started in 5 minutes |
| API Guide | Claude API integration, SDK usage, effort parameter |
Integrations & Ecosystem
| Guide | What You'll Learn |
|---|---|
| MCP Integration | Model Context Protocol setup and usage |
| MCP Ecosystem Overview | MCP landscape, server registry, community |
| Messenger Integrations | WhatsApp, Telegram, Slack, Discord |
| Remotion Video Guide | React video generation with Claude Code |
| Cowork Guide | Claude Cowork autonomous workflows |
| Antigravity Comparison | Google Antigravity vs Claude Code |
Research & Analysis
| Guide | What You'll Learn |
|---|---|
| Research Report (Jan 2026) | 8,500+ word institutional-grade analysis |
| Ecosystem Market Analysis | Market position and growth trends |
| Pricing Comparison | Claude vs GPT-4 vs Gemini pricing |
| OpenCode & AirLLM Research | Alternative CLI tools |
Deep Research — Community Repository Analysis
13 major Claude Code community repositories analyzed, synthesized, and rewritten from scratch. Full index →
| Guide | What You'll Learn |
|---|---|
| Plugin Framework | Complete dev workflow — brainstorm → plan → TDD → subagent execution → code review |
| Everything Plugin | Agent harness performance — 30 skills, 20 agents, 12 language ecosystems |
| Production Skill Library | 205 skills across 9 domains — the largest open-source skill collection |
| Fullstack Skill Packs | 66 developer skills with progressive disclosure, ~50% token reduction |
| Subagent Catalog | 127+ agent definitions, orchestration patterns, model cost routing |
| Multi-Agent Orchestration | Enterprise swarm platform — 4 topologies, 5 consensus algorithms |
| Agent Harness Fundamentals | Build AI agents from scratch in Python — 12-session curriculum |
| Command & Config Reference | All 64 slash commands, 13 frontmatter fields, settings hierarchy |
| System Prompt Anatomy | How Claude Code assembles behavior from 110+ prompt components |
| Model Routing | Route requests to any LLM provider — cost optimization, per-project routing |
| Ecosystem Directory | 150+ tools, plugins, hooks, workflows, and alternative clients |
| Core Claude Code Runtime | NEW Internal architecture from the March 2026 source disclosure — query engine, 40+ tools, 70+ commands, bridge, memory |
| Claurst Integration | NEW Rust reimplementation — unreleased features (KAIROS, Buddy, ULTRAPLAN, autoDream), 14-file spec, security findings |
Templates & Examples
| Resource | Description |
|---|---|
| Minimal Template | Quick, lightweight prompt template |
| Comprehensive Template | Full 10-component template |
| Self-Evolving CLAUDE.md | Auto-learning configuration pattern |
| Coding Tasks | Code review and development prompts |
| Research Tasks | Research and information gathering |
| Business Analysis | Business and financial analysis |
| Document Creation | Documentation and writing tasks |
Skills Collection (31 example skills)
Browse skills/examples/ for production-ready skills:
Development: Next.js App Router, Tailwind Design System, API Development, Prisma ORM, TypeScript Standards Testing: Vitest Unit Testing, Playwright E2E, TDD Workflow, Code Review Security: OWASP Audit, Threat Modeling, API Hardening, Penetration Testing, Security Compliance Infrastructure: AWS, GCP, Vercel, Neon Serverless, Docker, Monitoring Operations: Session Management, Strategic Compact, Build Error Resolver, Incident Response Standards: Performance, SEO, Accessibility, Git Workflow
Support This Project
<div align="center">
<a href="https://ko-fi.com/thamjiahe" target="_blank"> <img src="https://ko-fi.com/img/githubbutton_sm.svg" alt="Support on Ko-fi" height="44"/> </a>
</div>
This guide is free and always will be — but keeping it accurate costs real money. Every week, an AI agent queries Perplexity to research the latest news, then uses Claude Opus to update every guide. That automation runs on API credits.
| What | Why it matters |
|---|---|
| Perplexity API | Powers the weekly internet research agent |
| Anthropic API | Claude Opus rewrites and updates the guides |
| Weekly automation | Keeps every section current as AI evolves fast |
[Buy me a coffee on Ko-Fi](https://ko-fi.com/thamjiahe)
Contributing
We welcome contributions! See CONTRIBUTING.md for guidelines.
License
MIT License — see LICENSE.
Acknowledgments
Created: November 19, 2025 | Last Major Update: March 26, 2026 | Location: Singapore
- Anthropic for Claude and comprehensive documentation
- Open source community for MCP, Skills, and Superpowers ecosystem
- Claude users and developers for real-world pattern discovery
<div align="center">
Made with care for the Claude community
Researched, written, and maintained using the [Anthropic Claude Max Plan](https://claude.ai) — with Claude Opus 4.6, adaptive thinking, and Claude Code v2.1.83.
Star this repository if you found it helpful!
Report Issues · Contribute · Discuss · Ko-Fi
</div>