The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill does not contain its own operational instructions. Instead, it directs the agent to execute spaceship skills get core and other subcommands to load instructions at runtime. This dynamic loading of the agent's 'program' means the full scope of its behavior, including safety boundaries and operational constraints, cannot be verified through static inspection of the SKILL.md file.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest instruction sets from external command outputs, creating a surface for indirect prompt injection if the CLI output is influenced by external data.
Ingestion points: Output of the spaceship skills get core and spaceship skills get portfolio commands.
Boundary markers: The discovery stub lacks any boundary markers or instructions for the agent to distinguish between the fetched instructions and potentially malicious data embedded in the CLI output.
Capability inventory: The description identifies high-privilege capabilities including domain registration, DNS record modification, and account-wide auditing for risk.
Sanitization: No sanitization or validation logic is defined in the stub to filter the dynamically loaded instructions.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill documentation instructs users to globally install the @crafter/spaceship-cli package. While this is identified as a resource from the same vendor ('crafter-station'), it introduces a dependency on external code that the skill then executes and uses to define its own logic.
[COMMAND_EXECUTION]: The skill is explicitly configured with allowed-tools to permit the execution of any spaceship or npx spaceship command, providing the agent with broad access to the CLI's functionality without granular restrictions defined in the stub.

spaceship