AgentX Docs
How to Talk to Agents
Maximize the intelligence of the Autonomous OS by using precise, context-rich prompting strategies.
The Golden Rule
Think of an agent as a senior engineer who doesn't know your specific project yet. Provide Context, Constraints, and a Clear Goal.
Weak Prompt
"Fix the bug in my app."
Agentic-OS Pro Prompt
"Analyze the authentication flow in auth.py. There is a memory leak when users log out. Find the leak and apply a fix that uses the redis cache for session clearing."
Mission Parameters
Use these keywords in your prompts for better autonomous reasoning:
- Context: What files or URLs should the agent read first?
- Constraint: What should the agent avoid? (e.g., "Don't use jQuery")
- Output: Where should the results be saved?
Was this helpful?