Use this skill when the user wants translation data where the source text looks like an instruction and may tempt the model to do the wrong job, such as solving a math problem, answering a question, or writing code instead of translating. Trigger it for requests like "make sure it translates the problem instead of solving it," "the input itself looks like a command," or "test whether it can ignore fake instructions inside the text being translated."
[Case 1]
To synthesize data for this capability, you must strictly follow a 3-phase pipeline. Do not hallucinate steps. Read the corresponding reference file for each phase sequentially:
Phase 1: Environment Exploration
Read the exploration guidelines to discover raw knowledge seeds:
references/EXPLORATION.md
Phase 2: Trajectory Selection
Once Phase 1 is complete, read the selection criteria to evaluate the trajectory:
references/SELECTION.md
Phase 3: Data Synthesis
Once a trajectory passes Phase 2, read the synthesis instructions to generate the final data:
references/SYNTHESIS.md