| Titre | RTGS2017 NagaAgent 5.10 Path Traversal |
|---|
| Description | NagaAgent skill-management routes accept attacker-controlled skill names and directly concatenate them into filesystem paths without canonical path boundary checks. Specifically, user-provided name is used to create or delete directories under skill storage roots. Because path separators and traversal sequences are not rejected, crafted names can cause writes/deletes outside intended skill directories when the process has permission.
This behavior is in core API logic (not a helper-only dead path) and is reachable from:
- POST /skills/import
- DELETE /skills/{name} |
|---|
| La source | ⚠️ https://github.com/RTGS2017/NagaAgent/issues/311 |
|---|
| Utilisateur | CPT_Penner (UID 97246) |
|---|
| Soumission | 18/04/2026 20:17 (il y a 2 mois) |
|---|
| Modérer | 04/05/2026 18:01 (16 days later) |
|---|
| Statut | Accepté |
|---|
| Entrée VulDB | 360981 [RTGS2017 NagaAgent jusqu’à 5.1.0 Skills Endpoint extensions.py Nom directory traversal] |
|---|
| Points | 20 |
|---|