The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documentation and setup script recommend a pattern that pipes a remote script directly into a shell for installation.
Evidence: scripts/check_setup.sh and references/provider-operations.md include the command curl https://rclone.org/install.sh | bash.
[EXTERNAL_DOWNLOADS]: The skill fetches an installation script from the official and well-known rclone.org domain.
Evidence: https://rclone.org/install.sh referenced in scripts/check_setup.sh.
[COMMAND_EXECUTION]: The skill provides instructions and scripts that suggest the use of administrative privileges (sudo) to install software.
Evidence: scripts/check_setup.sh echoes installation commands using sudo bash and sudo apt install.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by retrieving and processing data (file names and remote metadata) from untrusted external sources.
Ingestion points: The skill executes rclone listremotes, rclone ls, and rclone lsd in scripts/check_setup.sh and throughout the SKILL.md workflow.
Boundary markers: Absent; remote names and file paths are not enclosed in delimiters or accompanied by instructions to ignore embedded commands.
Capability inventory: The skill has the ability to write to remote storage (rclone copy), delete files (rclone sync), and modify configurations (rclone config).
Sanitization: Absent; the skill does not explicitly validate or sanitize strings returned from remote storage listings before they are processed by the agent.

rclone