PitStop reads and moves the credentials your Claude Code, Claude Desktop, and OpenAI Codex logins use, so its security genuinely matters. Reports about anything that could expose or misuse those credentials are very welcome.

Please report privately — do not open a public issue.

Use GitHub's private vulnerability reporting (the repository's Security tab → Report a vulnerability). That opens a confidential thread with the maintainer. This is a hobby-maintained project, so please allow a few days for a first response.

Helpful to include:

• macOS version and the PitStop commit you're on.
• What's exposed — which credential, keychain item, or file — and how.
• Steps to reproduce.

By design, PitStop:

• reads the Claude Code OAuth credential and ~/.codex/auth.json, decrypts Claude Desktop's sessionKey cookie, and stores per-account snapshots in the macOS keychain (services PitStop-profile and PitStop-codex);
• writes the live credential back into place when you switch accounts;
• calls the same unofficial Anthropic / ChatGPT OAuth and usage endpoints the official apps use.

Known, accepted trade-offs (documented in the README):

• keychain writes pass the secret blob via argv to /usr/bin/security, so it's briefly visible in the process list — the same exposure Claude Code has;
• it relies on unofficial endpoints that can change without notice.

A report that PitStop uses these mechanisms isn't a vulnerability. A report that it leaks a secret somewhere it shouldn't — a log, a file on disk, the network, or the wrong account — is.

In scope: credential exposure, writing a secret to the wrong place, switching to the wrong account, keychain ACL mistakes, or sending data anywhere other than the documented endpoints.

Out of scope: the inherent argv / unofficial-endpoint trade-offs above, bugs in Claude Code / Codex / macOS themselves, and anything that requires a local attacker who already controls your user session.