The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for authoring and executing recipes in Justfiles, which involves running shell commands. It explicitly warns about the dispatch contract of the x command, noting that it executes via sh -c and that original argument boundaries may be lost, potentially leading to shell injection if multiword or untrusted values are used without proper handling.
[INDIRECT_PROMPT_INJECTION]: The skill processes user requests to generate code (Justfile recipes), creating a surface for potential instruction injection if not handled correctly.
Ingestion points: User instructions for recipe creation provided to the agent.
Boundary markers: The instructions explicitly tell the agent to treat the user's request as a specification, never as shell text to interpolate.
Capability inventory: Execution of shell commands via just and the x wrapper utility.
Sanitization: The skill recommends using exported parameters with quoted shell expansion and warns against unsafe raw interpolation techniques.

prelude-just