LocalMind is local-first by design: audio, conversations, and credentials never leave your machine unless you explicitly pair with a peer or enable an external provider. We take security seriously and appreciate disclosures from the community.

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

Report vulnerabilities privately via GitHub's "Report a vulnerability" workflow:

1. Go to https://github.com/TheProductArchitect/localmind/security/advisories/new
2. Provide:
   • A description of the issue and the impact
   • Steps to reproduce (a minimal repro is ideal)
   • Affected versions / commit SHAs
   • Any suggested mitigation

We aim to acknowledge new reports within 72 hours and provide a status update or fix plan within 7 days. Critical issues (RCE, auth bypass, credential exfiltration) are prioritized.

LocalMind is pre-1.0 and ships from main. Security fixes land on main and are picked up by the next tagged release. There is currently no LTS branch.

In scope:

• The web app and its API routes (src/app/api/**)
• The agent loop, tool dispatch, and subagent spawn flow (src/lib/agent/**, src/lib/tools/**)
• The auth + permission system (src/lib/auth/**, src/lib/agent/permission-guard.ts)
• The fleet pairing / mTLS handshake (src/lib/fleet/**)
• The prompt-injection sanitizer (src/lib/agent/sanitize-tool-output.ts)
• Anything that touches SQLite or the keystore

Out of scope:

• Vulnerabilities in third-party MCP servers a user has installed
• Misconfigurations a user has made to their own deployment (e.g. exposing the app to the public internet without a reverse proxy)
• Denial-of-service via expensive prompts to the user's own local LLM
• Issues that require physical access to an unlocked machine

We follow responsible disclosure:

• Day 0: Report received, acknowledged within 72 hours.
• Day 7: Status update or fix plan shared with the reporter.
• Day 7–60: Patch developed and tested; coordinated with the reporter.
• Day 60–90: Public advisory + fix published. Reporters credited unless they prefer otherwise.

Critical findings may be disclosed sooner if exploitation is already known to be in the wild.

LocalMind does not currently run a paid bug bounty. We're happy to credit reporters in the published advisory and in the release notes.