The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Installation of the Blacksmith CLI uses the curl -fsSL https://get.blacksmith.sh | sh pattern, which executes unverified remote scripts directly in the shell without integrity checks.
[REMOTE_CODE_EXECUTION]: The skill's primary functionality involves executing arbitrary commands on remote virtual machines via the blacksmith testbox run command.
[COMMAND_EXECUTION]: User-provided parameters such as <ID>, <command>, and <org-slug> are interpolated directly into shell command templates. This creates a risk of command injection if the agent does not properly sanitize these inputs.
[EXTERNAL_DOWNLOADS]: The skill downloads and executes scripts from get.blacksmith.sh, an external source that is not recognized as a trusted provider in this context.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface due to the lack of input handling for external data.
Ingestion points: User-supplied placeholders like <command>, <ID>, and <org-slug> in SKILL.md and repository content synchronized via rsync.
Boundary markers: Absent; instructions do not define delimiters or requirements for input validation.
Capability inventory: Execution of shell commands both locally (bun run testbox:run) and on remote environments (blacksmith testbox run).
Sanitization: No escaping, filtering, or validation is performed on the data before it is interpolated into executable commands.

blacksmith-testbox