rclone

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches installation assets from rclone.org, which is the official domain for the well-known rclone utility.
  • [REMOTE_CODE_EXECUTION]: Executes a remote shell script via curl piped to sudo bash to install the rclone utility, following the official software installation path.
  • [COMMAND_EXECUTION]: Utilizes system commands and the rclone binary to perform file operations and configuration. This includes the use of sudo for installation tasks.
  • [CREDENTIALS_UNSAFE]: Handles sensitive cloud provider credentials using placeholders in documentation to prevent accidental exposure during configuration.
  • [PROMPT_INJECTION]: The skill processes untrusted output from remote file listings, creating a surface for indirect prompt injection. * Ingestion points: Output from rclone ls and rclone lsd in SKILL.md and scripts/check_setup.sh. * Boundary markers: None identified to isolate remote data from agent instructions. * Capability inventory: File system and network operations via rclone (copy, sync, config). * Sanitization: No explicit sanitization or validation of remote file names or metadata before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 03:29 AM