The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of shell commands to manage the development environment and execute tests. This includes using bash to run generated scripts, python3 to execute model answers, and bunx to run platform-specific utilities.
[EXTERNAL_DOWNLOADS]: The skill uses bunx to fetch and execute the @exercode/problem-utils package. Additionally, it employs mise install to download and set up runtimes, package managers, and command-line tools as defined in the project's configuration files (e.g., mise.toml).
[DYNAMIC_EXECUTION]: The skill generates a temporary shell script (expected.sh) at runtime to automate the generation of expected output files from model answers. It also performs automated environment configuration by running mise trust --yes, which bypasses manual approval for tool-specific configuration files.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external "source material" and "learning objectives" to derive problem content. Since the skill has the capability to write files and execute shell commands, it presents an attack surface where malicious instructions embedded in the source material could potentially influence the agent's code generation or execution steps.

generate-judge-problems