**0 stars — Not suitable for developers who value transparency**

0 stars — Not suitable for developers who value transparency

I am extremely disappointed with the GitHub Copilot Harness. As an open-source developer, I need to understand and audit the environment in which an AI agent operates. Instead, the harness relies on hidden system prompts, opaque internal instructions, and restrictions that prevent the agent from disclosing the rules governing its behavior.

That is fundamentally counterproductive. If an AI tool is expected to work in my repositories, access files, run commands, and influence development decisions, I need full visibility into its prompts, policies, tools, permissions, and decision boundaries. Without that transparency, I cannot properly evaluate, trust, adapt, or integrate it into an open-source workflow.

The product is also too expensive for what it offers and does not feel competitive with more transparent alternatives. Rather than feeling supported as a developer, I feel misled and poorly served.

I will be cancelling my GitHub Copilot subscription.

Start with the GitHub Copilot app's /chat experience on Windows 10, using the reported app version 1.1.19 as context. The issue does not identify files, tests, or a specific implementation boundary; work would first need a defined transparency requirement and acceptance criteria.