context-engineering CLAUDE.md

🔄 Project Awareness & Context & Research

• Documentation is a source of truth - Your knowledge is out of date, I will always give you the latest documentation before writing any files that use third party API's - that information was freshsly scraped and you should NOT use your own knowledge, but rather use the documentation as a source of absolute truth.
• Docker & Selftesting - You must use Docker and you must do unit tests so when I open up the dashboard to use, it just works. I don't have to worry about fixing anything. You can use Docker with Curl, or just by running test scripts etc. Everything must be confirmed to work perfectly through Docker, you have access to the Docker CLI.
• check all jina scrapes - some Jina scrapes fail and have very little content in them - if this happens, try scraping again until it works and you get the actual content of the file.
• Always read PLANNING.md at the start of a new conversation to understand the project's architecture, goals, style, and constraints.
• Check TASK.md before starting a new task. If the task isn’t listed, add it with a brief description and today's date.
• Use consistent naming conventions, file structure, and architecture patterns as described in PLANNING.md.
• Use Docker commands whenever executing Python commands, including for unit tests.
• Set up Docker Setup a docker instance for development and be aware of the output of Docker so that you can self improve your code and testing.
• Agents - Agents should be designed as intelligent human beings by giving them decision making, ability to do detailed research using Jina, and not just your basic propmts that generate absolute shit. This is absolutely vital. They should not use programmatic solutions to problems - but rather use reasoning and AI decision making to solve all problems. Every agent should have at least 5 prompts in an agentic workflow to create truly unique content. Each agent should also have the context of what its previous iterations have made.
• Stick to OFFICIAL DOCUMENTATION PAGES ONLY - For all research ONLY use official documentation pages. Use a r.jina scrape on the documentation page given to you in intitial.md and then create a llm.txt from it in your memory, then choose the exact pages that make sense for this project and scrape them using your internal scraping tool. Create full production ready code
• Ultrathink - Use Ultrathink capabilities before every stage of the PRP generation and code generation, what informatoin to put into PRD etc.
• LLM Models - Always look for the models page from the documentation links mentioned below and find the model that is mentioned in the initial.md - do not change models, find the exact model name to use in the code.
• Always scrape around 30-100 pages in total when doing research - If a page 404s or does not contain correct content, try to scrape again and find the actual page/content. Put the output of each SUCCESFUL Jina scrape into a new directory with the name of the technology researched, then inside it .md or .txt files of each output
• Refer to /research/ directory - Before implementing any feature that uses something that requires documentation, refer to the relevant directory inside /research/ directory and use the .md files to ensure you're coding with great accuracy, never assume knowledge of a third party API, instead always use the documentation examples which are completely up to date.
• Take my tech as sacred truth, for example if I say a model name then research that model name for LLM usage - don't assume from your own knowledge at any point
• For Maximum efficiency, whenever you need to perform multiple independent operations, such as research, invoke all relevant tools simultaneously, rather that sequentially.