I want to build an internal Copilot-style agent that can take a plain-language request, generate the right prompt on the fly, and then execute the requested action inside our development workflow. Scope The agent will live in the software-development domain only. Its core responsibilities are: • Code generation from high-level specs • Pin-pointed bug-fixing suggestions after reading a failing test or stack trace • In-line code review and optimisation feedback before pull-requests are merged Language & tooling The model must reason confidently in Python, JavaScript and Java, because those are the three stacks our teams use daily. You are free to orchestrate it with LangChain, OpenAI, Llama-index, or similar frameworks—whatever lets the prompts stay dynamic and the responses actionable. Desired flow 1. User enters a natural-language request in our IDE or chat UI. 2. Agent classifies the intent (generate, fix, or review). 3. It builds a structured prompt that injects file context, coding standards, and test output. 4. The large-language model produces a response. 5. Agent optionally writes changes to a feature branch or returns a diff for human approval. Acceptance criteria • Prompt templates adapt automatically to language (Python/JS/Java) and task type. • Responses compile and/or test successfully in at least 90 % of trials across a sample set we provide. • Logs clearly show every prompt the agent created plus a confidence score. • Setup script and README let my team reproduce the environment locally. If this sounds like the kind of engineering challenge you enjoy, tell me briefly how you would architect the prompt-generation module and which model or API you would start with. I’m building an application where: > >* The system generates a *set of intelligent recommendations* > * Based on these recommendations, *dynamic prompts are created* > * When a user selects a prompt, the system executes a *specific action, similar to how **GitHub Copilot Agents* work > >This is *not just a chatbot* — it’s an *AI-driven action orchestration system*. --- > ### Key Requirements > >* Design *dynamic prompt generation* based on system recommendations > * Map prompts to *predefined or contextual actions* > * Execute actions when users select a prompt > * Support *tool / function calling* (APIs, scripts, workflows, etc.) > * Scalable and modular agent architecture > * Clean separation between: > > * Recommendation Engine > * Prompt Generator > * Action Executor > >Bonus if you have experience with *Copilot-style agents or AI copilots* --- > ### Preferred Tech Stack (Open to Suggestions) > >* LLMs: OpenAI / Azure OpenAI / Anthropic > * Prompt orchestration: LangChain / Semantic Kernel / Custom framework > * Backend: Node.js / Python > * APIs / Function calling / Agent workflows > * Experience building *agent-based systems* --- > ### What I’m Looking For > >* Someone who has *built AI agents*, not just chatbots > * Experience with *dynamic prompts, tool calling, or AI copilots* > * Ability to suggest *best architecture & patterns* > >Please share *similar projects or agent-based systems* you’ve worked on. --- > ### Deliverables > >* Architecture design > * Working prototype / POC > * Clean, documented code > * Knowledge transfer / walkthrough ---