Aspect Code logoAspect Code

Complaint Loop

Give feedback to improve AGENTS.md.

How complaints work

While Aspect Code is running in watch mode, you can type plain-English feedback directly into the terminal dashboard. An LLM agent reads your complaint and patches AGENTS.md accordingly.


Example complaints:

  • "The AI keeps putting new routes in lib/ instead of app/api/"
  • "It never runs migrations after changing models"
  • "Stop suggesting refactors of the auth module"

    • This requires an API key (see LLM Optimization).

    What happens when you complain

  • You type your feedback in the terminal dashboard
  • An LLM agent reads your complaint and the current AGENTS.md
  • The agent identifies which section(s) need updating
  • AGENTS.md is patched with new or updated rules
  • The change is shown in the terminal for your review

    • Complaints are a fast way to teach the system about patterns that static analysis alone can't detect — like team conventions, workflow preferences, or recurring AI mistakes.

    LLM OptimizationExisting AI Config