FIXING Opus 5: PROOF that Prompt Engineering IS NOT DEAD
Opus 5 is one of the BEST models ever released, and one of the WORST to talk to. Verbose, "load-bearing" everywhere, torching your output tokens. So we FIX it with the skill everyone said was dead: PROMPT ENGINEERING. 🔥 😲 Phase 3 Early Sign Up is Officially Available: https://agenticengineer.com/ia3-pre-launch?y=S_QdQ1G4GlU 🎥 VIDEO REFERENCES • Fix Opus 5 Codebase: https://github.com/disler/fixing-smartass-opus-5 • Software Factory (Advanced): https://youtu.be/haUfb1ievTE • Zucks Post: https://www.meta.com/thefutureisforeveryone/?srsltid=AfmBOoqCUQLgMU5sfTup2SoH9ZpFlQ_wwSMLzccQkoziF-mR5iiDExgI PUSH YOUR AGENTIC ENGINEERING BEYOND Tactical Agentic Coding: https://agenticengineer.com/tactical-agentic-coding?y=S_QdQ1G4GlU 🚀 In this video, IndyDevDan turns smartass Claude Opus 5 into a precise senior engineer with SYSTEM prompt engineering. There are two ways to prompt your AI coding agents: the user prompt (one task) and the system prompt (the LAW for every task). Most engineers fixate on the user prompt and skills. The real leverage is the system prompt, where every word you write gets multiplied over every single prompt in and every response out of Claude Code. 🛠️ Key takeaways: Positive + Negative Patterns: Tell your agent exactly what to replicate and exactly what to avoid (goodbye "load bearing", "worth stating plainly", em dash chains). Reference Points: Encode D1/R1/F1 style codes so you and your agent build a fast shared language and stop repeating tokens. Hard Operational Boundaries: Deliver ONLY what was requested. No scope creep, no surprise refactors, no co-author commit messages. Aliases: SCR, ELI, FOC, REF. Micro skills baked into your system prompt that expand into full instructions on demand. Examples (In-Context Distillation): Show real do/don't responses, even pulling responses from models you like, to lock in communication patterns. 💡 The system prompt is where prompt engineering pays off most because it applies across EVERY user prompt. Great engineering is great communication, and with agents, you communicate through prompts. Master the system prompt and you control your model's verbosity, cost, and quality, no matter which state-of-the-art model ships next. 🌟 Keep the smart, drop the ass. If you understand your tools, you understand the results you can get from them. Your system prompt is the law for your agents. Use it. Stay Focused and Keep Building. IndyDevDan 📖 Chapters 00:00 Opus 5: The Best Worst Model Ever Released 01:53 System Prompt Engineering 02:49 Fixing Smartass Opus 5 04:43 System Prompt BEATS User Prompt 08:06 Positive AND Negative Prompt Engineering 13:22 Reference Points 16:44 Hard Operational Boundaries 18:38 Prompt Aliases 26:40 Example Distillation #promptengineering #agenticengineering #claudecode
