I Didn't Know Where My Keys Had Leaked, So I Started Building GuestSafe
After a compromised Mac and AI agents leaked secrets into session logs, I began building GuestSafe: a local, auditable alternative to .env.local.
#security
4 articles
Local-First Secrets for AI Agents on macOS: A 2026 Field Survey
A field survey of local secret stores for AI agents on macOS—and the privilege boundary needed to keep an agent from erasing its own access trail.
Security Analysis of Claude Code v2.1.88 — Source Reconstructed from Source Maps
A deep-dive security review of Claude Code's internals, reconstructed from its published source maps. Covers permission sandboxing, shell execution, credential handling, remote killswitches, and 10 findings.
Address Poisoning is a Wallet Bug, Not a User's Fault
Address poisoning isn't a user problem — it's a wallet UI bug. How wallets enable this $68M attack and how to fix it.