Prompt
Design a skill to use when repeating the task that caused the failed test. [input] AI and execution environment to use: Project skills used in Codex Work materials and examples: Failure log and test code Conditions to maintain and scope of work: Written in Korean and composed of only the files actually needed. Distinguish between executed verification and non-executed verification. [Specialized tasks for this Skill] Distinguish between product defects, incorrect expectations, environment, and indeterminacy. Determine the steps necessary for this task among data collection, judgment, result writing, and verification, and separate the inputs that change every time and the rules to be fixed. Give priority to facts confirmed in the current environment and provided code. Don't create APIs/files/tests that don't exist. Treat sensitive values as hidden examples and leave unexecuted tasks as unexecuted. [File for registration] Please write the entire SKILL.md that can be saved immediately. In the name and description, indicate what to do and when to select it, and in the body, include input, procedure, result format, exception handling, and completion criteria. Separate long cases or forms into references or assets only when actually necessary, and connect the reading points. Don't add unnecessary scripts or dependencies to simple tasks. [Confirmation and registration] Provide examples of normal requests, missing information, or non-applicable requests and expected behavior. Please explain the installation path and calling method based on official instructions or confirmed settings for the usage environment. Distinguish between those that you wrote, installed, and executed yourself and those that only suggest content, and leave unverified items.