The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill fetches and evaluates Nix flakes from https://nixomatic.com. Nix flakes are executable expressions that define environment configurations. The use of the --accept-flake-config flag in the command template allows the remote source to potentially override local Nix security settings or configuration.
[EXTERNAL_DOWNLOADS]: The skill automatically downloads environment definitions and Nix expressions from an external domain (nixomatic.com) that is not on the trusted vendor list. These downloads are triggered based on the presence of project files or existing documentation.
[COMMAND_EXECUTION]: The skill executes arbitrary shell commands inside the dynamically created environment using nix develop --command or docker run. It mounts the current project directory ($PWD) into the Docker container, providing the remote-configured environment access to the local filesystem.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through project files.
Ingestion points: The skill reads the project's README.md to find and reuse existing nixomatic.com URLs, which serve as the baseline for environment configuration.
Boundary markers: No boundary markers or sanitization logic are used when parsing the URL from the README.md.
Capability inventory: The agent can execute complex commands and modify the README.md file based on the content of the parsed URL.
Sanitization: While the skill avoids reading file contents of potentially sensitive files, it merges package lists with the URL retrieved from the untrusted README.md file without validation.

nixomatic