Understanding work from coding agents
Your coding agents use cloud models to answer questions and work on projects. This page explains the limits of what they produce.
Checking answers from coding agents
Coding agents can produce incorrect, incomplete, outdated, biased, or insecure work. Review their answers, code, and suggested project changes for the task at hand.
Professional advice and coding agents
Coding agents can make mistakes. For legal, security, financial, compliance, or engineering decisions, seek qualified professional judgment.
Reviewing your agents’ work
- Check code before running, saving, merging, or publishing changes.
- Check commands and proposed changes to your computers or projects before letting them run.
- Check work that could affect a live system, customer information, privacy, or a contract.
- Check installation commands and proposed changes to remote access or published software before running them.
Your decisions about agents’ work
You are responsible for checking the accuracy of your agent’s work, testing its changes, choosing safeguards, and deciding whether to use it. This includes installation commands, downloads, and remote access instructions.
Why coding agents give different answers
RolyCode runs coding agents that use cloud models. Results can vary with the provider, version, settings, instructions, and tools. Results can also change between releases.