package-k3s-green

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The green launcher script fetches and executes code from the vendor's repositories (github.com/getcolors/k3s.git, github.com/getcolors/once.git, and github.com/getcolors/green.git) at runtime using the babashka.deps/add-deps function.
  • [DYNAMIC_EXECUTION]: The script employs dynamic loading techniques to execute functionality by using requiring-resolve to find and invoke functions from the downloaded dependencies based on user-supplied commands.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes content from a public Git repository specified in the colors.yml file, creating a surface for indirect prompt injection.
  • Ingestion points: The Git repository specified in colors.yml.
  • Boundary markers: No explicit delimiters are used to prevent the agent from treating repository data as instructions.
  • Capability inventory: The skill can execute kubectl commands with sudo privileges, run OpenTofu for infrastructure changes, and Ansible for configuration.
  • Sanitization: The provided instructions do not detail any sanitization of the repository content before processing.
  • [COMMAND_EXECUTION]: The skill is designed to run sensitive shell commands, including sudo k3s kubectl on the target VPS, providing broad cluster administrative access.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 01:58 PM