𝗔𝗜 𝗪𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀 𝗙𝗼𝗿 𝗟𝗮𝗿𝗴𝗲 𝗖𝗼𝗱𝗲𝗯𝗮𝘀𝗲𝘀

𝗔𝗜 𝗪𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀 𝗙𝗼𝗿 𝗟𝗮𝗿𝗴𝗲 𝗖𝗼𝗱𝗲𝗯𝗮𝘀𝗲𝘀

Large codebases are hard. My project has 75,000 lines. I do not hold it all in my head. You likely face this too.

LLMs predict the next word. They do not see the world like you. They lack your life experience. You must provide the context.

LLMs mirror your tone. If you are skeptical, the AI is too. Stay factual. State goals plainly. Use specs to avoid bias.

Stop searching for a better prompt. Use a system.

• /grill-with-docs: Set the context. Use a CONTEXT.md file. Use ADRs for hard decisions.
• /to-prd: Turn context into a plan.
• /to-issues: Break the plan into small tasks.
• /tdd: Write code using tests.

This is Context Engineering. Spend effort up front. Set durable constraints. This stops the AI from drifting.

Context is the key. Provide it clearly. Let the AI do the work.

Source: https://dev.to/malcolmmcneely/ai-agentic-workflows-on-large-codebases-3275 Optional learning community: https://t.me/GyaanSetuAi