The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the KASM Workspaces installer from the software vendor's official distribution point on Amazon S3 and fetches the cloudflared binary from Cloudflare's official GitHub repository.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands and scripts requiring administrative privileges to perform system-level tasks such as installing Docker, configuring swap files, managing KASM services, and running database maintenance operations. It also provides instructions for remote command execution via SSH.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection within its automation scripts. Several environment variables intended to be supplied by the user are interpolated directly into shell commands without validation or escaping, which could lead to command injection if malicious strings are provided.
Ingestion points: Environment variables such as CLOUDFLARE_API_TOKEN, TUNNEL_NAME, and TUNNEL_HOSTNAME in references/cloudflare-tunnel.md and the assets/env-template file.
Boundary markers: There are no delimiters or specific instructions used to distinguish user-provided data from command-line logic.
Capability inventory: The skill's scripts utilize high-impact capabilities including network operations via curl, remote shell access via ssh, and container management via docker.
Sanitization: There is no evidence of input validation, shell escaping, or other sanitization measures for external data before it is executed in a subshell environment.

kasm