Update to OceanToken

{
  "description": "Choose, compare and price AI models on OceanToken, and hand sub-tasks to other models through it. Use when the user asks which model to use, wants prices, a cheaper alternative or a cost estimate, asks for a second opinion from another model family (GPT, Claude, Gemini, DeepSeek, Qwen, Llama...), or has bulk text work (summaries, translation, classification, test data, docs) that a cheaper model can do. Also for Chinese requests such as 选模型、比价、哪个模型便宜、估算费用、要花多少钱、换个模型、第二意见、交给便宜的模型.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 270
    }
  ],
  "name": "oceantoken-models",
  "skill_md_contents": "---\nname: oceantoken-models\ndescription: Choose, compare and price AI models on OceanToken, and hand sub-tasks to other models through it. Use when the user asks which model to use, wants prices, a cheaper alternative or a cost estimate, asks for a second opinion from another model family (GPT, Claude, Gemini, DeepSeek, Qwen, Llama...), or has bulk text work (summaries, translation, classification, test data, docs) that a cheaper model can do. Also for Chinese requests such as 选模型、比价、哪个模型便宜、估算费用、要花多少钱、换个模型、第二意见、交给便宜的模型.\n---\n\n# Choosing, pricing and delegating with OceanToken\n\n## Find and compare\n\n- `search_models` filters by `type`, `publisher`, `capabilities` (`vision`,\n  `tools`, `reasoning`, `web_search`, `audio_output`) and `max_price`, and\n  `sort=\"price\"` puts the cheapest first. Search short names (\"veo\", \"opus\",\n  \"qwen\") and read ids from the results; never guess an id.\n- Prices are in the unit the model is billed in: chat per 1M input / output\n  tokens, images per image (or image tokens), video per second (a \"from\" price:\n  resolution and audio cost more), speech per 1K characters, transcription per\n  minute.\n- When comparing, show a short table: model, price, context window, relevant\n  capabilities, and one line on why each fits.\n\n## Estimate before spending\n\n`estimate_cost` returns USD, a `confidence` (`close`, `rough`, `minimum`,\n`unknown`), the assumptions made and whether it fits the available balance.\nPass real numbers when you have them: `input_text` or `input_tokens`,\n`output_tokens`, `seconds` + `resolution` for video, `text` for speech. Report\nthe estimate with its confidence; do not present a `rough` or `minimum` figure\nas exact.\n\n## Delegate with ask_model\n\n`ask_model(model, prompt, ...)` sends one self-contained request to another\nmodel and returns its answer, usage and cost.\n\n- Good uses: bulk or mechanical text (summarise 30 files, translate strings,\n  classify tickets, draft test fixtures) on a cheap model; a second opinion from\n  a different model family on a design or a diff; a long-context model for a\n  document too big to reason over comfortably; a vision model for screenshots\n  (`image_urls`).\n- The delegate sees nothing but your prompt: include the code, the context and\n  the exact output format you want.\n- Check what comes back. Treat it as a draft from a colleague, not as truth, and\n  say in your reply that another model contributed.\n- Say what it cost when it is not trivial (`cost_usd`).\n\n## Balance\n\n`get_account` shows balance and available balance (minus holds for running\nvideo jobs). If a request is refused for balance, tell the user that credit is\nadded in the OceanToken console, and offer a cheaper model.\n"
}