Mnemos is pre-1.0. Security fixes land on the latest minor release only.
Upgrade to the newest 0.x.y before reporting a vulnerability — if the
issue is already patched there, an upgrade is the fix.
| Version | Supported |
|---|---|
| 0.3.x | ✅ (latest minor) |
| < 0.3 | ❌ |
Please do not open a public GitHub issue for security reports.
Use GitHub's private vulnerability reporting instead:
- Go to https://github.com/mnemos-dev/mnemos/security/advisories/new
- Describe the issue, reproduction steps, and the impact you observed.
- Include the mnemos version and backend (chromadb / sqlite-vec).
You should get an acknowledgement within 7 days. Fix-in-flight and disclosure timing are negotiated in the advisory thread.
Mnemos runs locally, reads your Obsidian vault and Claude Code JSONL
transcripts, and (optionally) calls the Claude API via mnemos-dev[llm].
Security-relevant areas:
- Path traversal / write-outside-vault — mining or refine paths escaping the configured vault root.
- Arbitrary command execution — the auto-refine hook spawns
claude --print. Injection into that command line. - Secret leakage — API keys or tokens accidentally written to
drawers, logs, or the
.mnemos-hook-status.jsonfile. - Denial-of-service on the user's own machine — a crafted transcript
or
mnemos.yamlthat hangs the miner or fills disk.
- Obsidian itself, ChromaDB, sqlite-vec, or Claude Code — report those upstream.
- Issues that require already-root/admin on the machine.
- Anything that depends on the user intentionally installing a hostile plugin or skill.