The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's core functionality relies on the generation and execution of shell commands through the claude CLI. It provides templates for spawning background and interactive processes using the --print flag to pass complex multi-line instructions.
[PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection. User-supplied goals and tasks are interpolated directly into the instructions given to spawned sub-agents without clear sanitization or boundary protection.
Ingestion points: User-defined tasks and goals provided to the /spawn and /spawn-manager commands.
Boundary markers: The templates do not define clear delimiters (e.g., XML tags) or instructions to ignore embedded commands within the user input.
Capability inventory: Sub-agents are granted capabilities such as file modification (Edit, Write), repository access (Bash), and system discovery (list_peers).
Sanitization: There are no explicit instructions for the coordinator to escape or validate user-provided strings before they are passed to the shell.
[EXTERNAL_DOWNLOADS]: The documentation recommends the use of an external third-party utility, claude-squad, hosted on GitHub (github.com/smtg-ai/claude-squad). This introduces a dependency on software from a non-trusted vendor for monitoring operations.

spawn-manager