keygrant blogNotes on keeping secrets out of AI agents' context.https://keygrant.app/Claude Code keeps printing my secrets: why hook redaction isn't enoughhttps://keygrant.app/blog/hook-redaction-isnt-enough/https://keygrant.app/blog/hook-redaction-isnt-enough/Agents leak API keys through output, tool-call arguments and generated code. Redacting output with hooks treats the symptom. Here's the architectural fix, and the base64 bug I found in my own redaction.Sat, 10 Oct 2026 00:00:00 GMT