When Your Accommodation Becomes a Credential Leak
I use Claude Code every day. Autistic, ADHD – executive function gaps mean I ramble, lose threads, start multiple essays while finishing none. AI helps shape what I'm trying to say into something a reader can follow. I've written about this: the ideas, the words, the stories are mine. The thread is ours. It's accommodation, not deception. November 5, Apiiro security researchers published findings that made me go back through every project I've touched: Claude Code's "approve and don't ask again" mechanism has been leaking credentials into public GitHub repositories. 1,379 distinct secrets across 1,122 files, traced from 104,185 scanned copies of .claude/settings.local.json on public GitHub (Apiiro, November 5, 2026; scans conducted November 1-3, 2026; https://www.qodo.ai/blog/claude-code-dont-ask-again-secrets/ ). 379 database passwords. 43 SSH logins. 42 GitHub/GitLab/npm tokens. 23 AWS keys. All sitting in plain text in settings files developers com...