Claude Code leaks over 1,300 secrets into public GitHub repositories

Claude Code is Anthropic’s development tool for working with code, and the issue here sits in the files and workflows that developers use around it. GitHub repositories are a common place to store project material, so exposed secrets in that environment matter because they can be reused after they are written down. The appearance of both repository leaks and approval-file retention points to the same operational layer, where tooling meets day-to-day development practices.

Developers who rely on Claude Code now have to treat saved approvals and repository hygiene as part of their security setup, not as afterthoughts. Passwords and tokens in those files can outlive the moment they were created, which makes cleanup and access control more important for teams that use shared code hosting. Built-in secret scanning, if adopted, would add another check before sensitive data lands in places that are easy to copy or publish. (per thedefiant.io)

The tension is between fast AI-assisted coding and the old problem of secret sprawl, where convenience leaves credentials scattered across files and services. Claude Code is not the only tool in that category, so this case reads as a warning about how quickly new developer workflows can widen the surface area for credential exposure. Security features that catch secrets earlier can reduce exposure, but they also show how much trust these tools place in users to manage sensitive data correctly.

Still open is how broadly the exposed material was distributed, who could access the affected repositories or approval files, and whether the proposed safeguards will be turned on in future Claude Code releases. Anthropic has not yet said which secret-scanning protections will ship or when they will appear. The practical test is whether the next version of the tool blocks or flags the same kinds of password and token storage before they reach GitHub.

The total crypto market barely moved in the 4 hours after the first headline, after falling 0.3% in the 2 hours before.

42 prices, one every 30 minutes · The move was larger before the first headline than after: an observation, not an explanation. The market also moves on things nobody publishes.