obsidian-tc memory tools can overwrite and corrupt vault notes
An AI assistant using this Obsidian tool can accidentally erase the text of a note that it does not own, because two different memory items can map to the same file. Failed operations could also leave the memory data out of sync with the no
| Severity | HighCVSS 7.1 of 10 |
|---|---|
| Being exploited? | UnclearNo reports either way yet |
| Fix available? | Yes1.31.4 · since 25 Oct 2026 |
| How sure is this? | Official, 2 sourcesBacked by an official source |
How the attack works simplified
- 1Agent writes memoryCaller uses create or rename memory tool
- 2Path computedmemory/<type>/<name>.md path derived
- 3Note overwrittenExisting foreign note body replaced
- 4DB out of syncRow changed before failed write
- ResultUser note content lost and memory inconsistent
What happened
In obsidian-tc's memory lifecycle tools (create_entity, link_entities, rename_entity, unlink_entities, delete_entity), memory notes at memory/<type>/<name>.md are regenerated without an ownership check, so a hand-written or foreign note at that path can have its body overwritten. Some tools also change SQLite before the file write, so a refused write can leave the database inconsistent.
Are you affected?
You are affected if you run obsidian-tc version 1.0.1 through 1.31.3 with the memory tools (write:memory scope) enabled.
| Package / product | Affected versions | Fixed in |
|---|---|---|
| obsidian-tc · npm | >= 1.0.1, <= 1.31.3 | 1.31.4 |
What to do
- Update obsidian-tc to version 1.31.4 or later
- Back up your Obsidian vault, especially the memory/ folder, before updating
- If you cannot update yet, disable the write:memory scope for AI agents so the memory write tools cannot run
Timeline
- 23 Oct 2026, 15:17 UTCPublicly disclosed
- 23 Oct 2026, 15:17 UTCReported by NVD · NVD
- 23 Oct 2026, 18:47 UTCReported by GitHub Advisory · GitHub Advisory
Sources 2 reports, oldest first
Official vendor, government CERT or advisory database
- NVD23 Oct 2026, 15:17 UTC
- GitHub Advisory23 Oct 2026, 18:47 UTC
Technical details
- ID
- CVE-2026-115478 · GHSA-r3g7-r4hr-m258
- CVSS
- 7.1 / 10
- EPSS
- 0.3% chance of exploitation within 30 days
- Disclosed
- 23 Oct 2026, 15:17 UTC
- On Zero Radar
- before the radar started
- Last change
- 25 Oct 2026, 13:32 UTC
The plain-language parts and the attack diagram were written by an AI model (anthropic/claude-haiku-5.5) from the sources above. Facts like versions, exploitation and dates come from the sources themselves. How Zero Radar works