The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill contains installation commands for the nb-cli utility (curl -sSL ... | sh and irm ... | iex) that target the vendor's GitHub repository. It minimizes risk by strictly forbidding the agent from executing these commands itself, ensuring that software installation remains a manual user-initiated process.
[COMMAND_EXECUTION]: The skill executes the nb-cli command-line tool to manage cloud resources. It employs a highly defensive workflow that mandates the use of --help for command verification, requires a --dry-run for all mutating actions, and necessitates explicit user approval for each individual command that changes resource state.
[EXTERNAL_DOWNLOADS]: The skill references binary downloads and update paths hosted on the author's GitHub repository (CLOUDFOREST-CO-TH/nb-cli-releases). These resources are expected for the skill's functionality and are sourced from the vendor's own infrastructure.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it processes and displays data retrieved from the nb-cli utility.
Ingestion points: Data and notices returned from subcommands such as list, get, and version (SKILL.md).
Boundary markers: No explicit delimiters are used; the agent is instructed to relay CLI output verbatim to the user.
Capability inventory: The skill can execute powerful infrastructure commands like create, delete, and os install (SKILL.md).
Sanitization: The skill relies on the human-in-the-loop approval process and mandatory dry-run output review to mitigate risks from potentially malicious data in the CLI output.

nb-cli