maru manages credentials-adjacent state for AI coding agents. We take vulnerabilities seriously even though the project is pre-1.0.

Once 1.0 ships, we'll commit to fixing the latest minor and one prior.

Please do not open a public GitHub issue for security reports.

Email security@maru.dev (or the maintainer listed in Cargo.toml's authors field if that address isn't yet reachable). Include:

• A clear description of the issue.
• Steps to reproduce, or a proof-of-concept.
• The version and platform you saw it on.
• Any mitigation you've already identified.

If you'd prefer encrypted email, ask in your first message and we'll exchange keys.

• Acknowledgement within 48 hours. If you haven't heard back, ping again — assume the message got lost.
• Triage within 7 days. We'll confirm the issue, assess severity, and propose a fix timeline.
• Fix and disclosure within 90 days for confirmed issues, sooner for severe ones. We'll coordinate disclosure timing with you.

In scope:

• Anything that causes maru to leak credentials (.credentials.json, auth.json, oauth_creds.json, OAuth tokens, API keys).
• Profile cross-contamination — one profile reading another's state without the user's intent.
• Privilege escalation, arbitrary file write outside $MARU_HOME, or arbitrary command execution from a .maru file or profile import.
• Anything that bypasses the GENESIS §8 deny-list during clone/export.

Out of scope:

• Bugs in the underlying agent CLIs (Claude Code, Codex, Gemini). Report those upstream.
• Issues that require an attacker to already have your local user account.
• Theoretical issues without a working repro.

If you'd like public credit, we'll list you in the CHANGELOG.md entry for the fixing release. Anonymous reports are also fine.