Please report privately, not via a public issue:

• GitHub → the repository's Security → Report a vulnerability (private advisory), or
• email contact@seanfloyd.dev.

You'll get an acknowledgement within a few days.

ccpool runs locally against your own ~/.claude data. The surface worth a security report:

• ccpool init writes/merges ~/.claude/settings.json — a bug that clobbers, corrupts, or injects unintended hooks/commands into that file (it's meant to be idempotent, never-clobber, symlink-aware, and to back up before writing).
• The hooks/statusline path executes as part of Claude Code — anything that could let crafted input (a malformed payload on stdin, a poisoned snapshot/transcript file) cause code execution, a crash that breaks Claude Code, or a write outside the expected paths.
• ccpool run sets subagent model/effort env and execs a command — anything that could let it run an unintended command.

Out of scope: the accuracy of the $/pace estimates (they're advisory, API-equivalent, and self-calibrated — see the README's "Honest limitations"), and issues in ccusage itself (report those upstream).