The Agent Skills Directory

[SAFE]: The skill is composed entirely of Markdown instructions and JSON evaluation cases. It does not include any executable scripts, shell commands, or binary assets.
[DATA_EXPOSURE]: Although the instructions encourage the agent to inspect the local repository for context (e.g., reading code or documentation), there are no mechanisms for data exfiltration or access to sensitive credential files like .env or .ssh.
[COMMAND_EXECUTION]: The skill configuration explicitly includes disable-model-invocation: true, which prevents the model from calling external tools or executing shell commands within this context.
[PROMPT_INJECTION]: The instructions are well-structured and focus on maintaining a specific 'interviewer' persona. There are no attempts to bypass safety filters, extract system prompts, or override agent constraints.

grill-me