idrive-backup
Pass
Audited by Gen Agent Trust Hub on Jul 19, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill accesses sensitive application configuration files in
~/Library/Application Support/IDriveforMac/. Specifically, it readsIDLoginDetails.plist, which contains the user's account email address, server IP address, and storage quota information, andappDefaultSettings.plist, which contains detailed file system exclusion maps. While the code does not include network exfiltration, these details are exposed to the agent context. - [COMMAND_EXECUTION]: The skill relies on a bundled Python script (
scripts/idrive_tool.py) to perform its primary functions. This script executes file system operations including reading and writing to XML plists, managing directory backups, and deleting configuration keys. While restricted to the iDrive configuration directory, it performs direct modifications to the application's runtime environment. - [PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection through its log analysis features. It parses structured XML (
LOGXML) and plain-text session logs generated by the iDrive application. Malicious content within file names or paths backed up by the system could be interpreted as instructions when the agent summarizes or diagnoses these logs. - Ingestion points:
~/Library/Application Support/IDriveforMac/SessionLogsNew/LOGXML/and~/Library/Application Support/IDriveforMac/SessionLogsNew/Backup/. - Boundary markers: The script does not employ specific delimiters or boundary markers to isolate log data from instructions.
- Capability inventory: The skill possesses the capability to modify backup sets and system-level application settings via the bundled tool.
- Sanitization: There is no evidence of sanitization or escaping applied to log content before it is presented to the agent for processing.
Audit Metadata