refine

# Refine Preserve the prompt’s intent, language, scope, audience, permissions, and explicit output contract. Rewrite the prompt; never execute it. ## Providers and sources Exactly two switches, with model names used only as documentation baselines: | Switch | Bundled guidance | Baseline and official documentation | |---|---|---| | `--for openai` | `prompts/openai.md` | GPT-6 Astra: [prompting best practices](https://developers.openai.com/api/docs/guides/latest-model#prompting-best-practices), [prompt engineering](https://developers.openai.com/api/docs/guides/prompt-engineering) | | `--for claude` | `prompts/claude.md` | Claude Fable 5.1: [Fable guide](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-fable-5-1), [Claude best practices](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices) | This table is drift-prone: documentation URLs move, model baselines are superseded, and the bundled guides go stale without announcing it. The freshness claim is therefore one line, on its own, in exactly this shape, and it covers both URLs, both baselines, and both bundled guides: ```text Checked: 2026-09-15 ``` Compare that date before promising the guidance is current. If it predates the release these skills shipped with, or either URL no longer resolves, the guidance is unverified: re-read both official pages, refresh the bundled guide deliberately, and update the line in the same edit. K...