Trainer Train - Prompt

Use this skill as the orchestration contract for one trainer run against a prompt-like target: any *.prompt.md, *.prompty, *.instructions.md, system prompt, or natural-language instruction file.

Read references/prompt-loop-contract.md for the full routing table, judge-mode rules, and prompt-specific validation constraints before any stage execution.

When to use this skill

• The selected target is a prompt file, instruction file, or prompty artifact.
• The caller needs to initialize or resume a trainer workspace for a prompt target.
• Missing datasets or eval manifests need to be synthesized before optimization.
• The optimization stage returns a manual follow-up payload and the loop must continue.
• A winning candidate needs to be validated and written back to the source prompt file.

Do not use this skill for code files, skill files, or agent contract files. Read the parent trainer skill's references/target-routing.md to identify the appropriate specialist for those target types.

Required inputs