Executive brief
Claude HUD, a terminal-based dashboard for the Claude Code tool, contains a security flaw that allows an attacker to read sensitive files on the host system. By providing a specially crafted file path through the application's input, an attacker can bypass directory restrictions to access data they should not be able to see. Additionally, the application records these unauthorized access attempts in a persistent cache file with weak security settings, potentially leaving a permanent record of the sensitive data on the disk.
Technical details
A path traversal vulnerability exists in Claude HUD through version 0.0.12 due to insufficient validation of the 'transcript_path' value received via stdin JSON. An attacker with local access can provide a manipulated path (e.g., using '../' sequences) to read any file accessible by the process user. Furthermore, the application writes file metadata (path, size, and mtime) to a persistent cache file with insecure permissions (exceeding 0o600), which allows the record of accessed paths to persist after the process terminates. The vulnerability was addressed in commit 234d9aa by canonicalizing transcript cache paths and enforcing restrictive file permissions.
Affected products
- jarrodwatts Claude HUD through 0.0.12
Timeline
- 2026-04-19: disclosed: Initial report by researcher Katriel Moses
- 2026-04-22: patched: Fixed in commit 234d9aa and PR 487
- 2026-05-18: advisory: CVE published to NVD