The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided Justfile content and natural language requests to generate shell commands.
Ingestion points: The agent reads existing Justfile recipes and user-defined specifications to author or edit tasks, as described in SKILL.md.
Capability inventory: The skill guides the execution of shell commands via just and the x utility, specifically for testing and verifying new recipes.
Boundary markers: Instructions explicitly tell the agent to treat requests as specifications rather than direct shell text to interpolate, which helps mitigate injection risks.
Sanitization: The skill mandates safer practices, such as using quoted shell expansion ("$value") for multiword or untrusted values and warns against raw text interpolation.
[REMOTE_CODE_EXECUTION]: The skill references a command that executes content from a remote repository via Nix.
Evidence: nix run github:darkmatter/prelude#skill -- guide command-conventions is mentioned in the references section of SKILL.md.
Context: This reference points to the official repository of the skill's author/vendor (darkmatter) for additional documentation.
[COMMAND_EXECUTION]: The skill involves creating and verifying shell-based recipes, which is its primary purpose.
Evidence: The skill provides instructions for using just --dry-run and x to verify command outputs and warns about the evaluation of backticks during Justfile loading.
Context: These instructions are defensive in nature, providing the agent with tools to safely inspect commands before execution.

prelude-just