Show HN: Confessor – replay what private info Claude Code accessed on your PC (github.com)

A new tool called Confessor has been unveiled to help users understand what actions an AI coding agent, such as Claude Code, performed on their computer. By analyzing session logs stored locally, Confessor reconstructs the entire history of interactions, including file access, commands executed, and any sensitive information that may have been exposed during an AI session. This is particularly significant for the AI/ML community, as it addresses growing concerns around privacy and data security when using powerful AI tools that have access to sensitive information like API keys or personal documents. Confessor stands out because it requires no installation alongside the AI agent and makes no network calls itself, ensuring that the tool is trustworthy and doesn't further expose user data. It identifies critical actions such as sensitive files being read followed by external network calls, thereby highlighting potential data exfiltration risks. With features that analyze logs from various AI services, including ChatGPT and Gemini, and relying on a rigorous detection engine based on predefined rules rather than machine learning, Confessor empowers users to audit their interactions retrospectively and make informed decisions regarding their data security.