Update to Givon AI

{
  "description": "Compare generation models when the user asks for a comparison or the trade-offs. Not needed when a valid model choice and current guidance are already in context.",
  "included_files": [],
  "name": "model-selection-guidance",
  "skill_md_contents": "---\nname: model-selection-guidance\ndescription: \"Compare generation models when the user asks for a comparison or the trade-offs. Not needed when a valid model choice and current guidance are already in context.\"\nmetadata:\n  version: skill-render-ea6bd93365667f6d\n---\n\n# Comparing models\n\nCompare only compatible options from the current catalog. The catalog owns availability,\ninputs, parameters, limits and price; this skill owns the criteria for a recommendation.\n\n## Pick the right action\n\nUse `givon_list_models` (CLI: `givon model list`) for browsing and comparison. Choose from\ncurrent active models by the requested result, reference media and priorities. Read the exact\nchosen model with `givon_get_model_schema` (CLI: `givon model get <id>`) before authoring\nits input. Use the catalog's `promptAuthoring` and schema `x-promptAuthoring` metadata for\nlanguage and reference semantics, together with the relevant craft and model-family guide.\nIf a current choice and guidance already cover the same inputs and constraints, reuse them.\nThe caller writes the final prompt; there is no separate selection-advisor or prompt-preparation\ntool on this surface. Finish through `givon-ai-create` when the user asked to generate; a\ncomparison or prompt-only request does not authorize a paid job. Treat an unknown capability\nas unknown and a lack of a compatible model as a constraint, not a service outage.\n\nFor a model the user named, check its contract; do not start a new contest between models.\nIf a current choice and the skills you need are already in context, use them while the\nrequirements stay the same. A change of inputs, operation or significant constraints\nrequires rechecking the affected facts; editing the text alone does not require reading the\nsame catalog or skill again.\n\nTell a request to choose apart from a request to make. For «сгенерируй фотореалистичного\nчеловека», choose a compatible model yourself and continue preparing the generation. Do not\nadd a separate reply that explains the choice or asks «какую модель предпочитаете?» when you\nhave enough data. Explain the choice when the user asks, or when a significant limit or\ntrade-off needs their decision. Preparation does not replace the price check or the user's\npermission to launch.\n\n## Comparison method\n\n1. Define the required result and the required inputs: text, a start frame, a reference,\n   a source video, text to be spoken or other material.\n2. Pick out the criteria that matter to the user: quality, speed, cost, editing, exact text,\n   identity, sound or another explicit priority.\n3. Compare the candidates by purpose, inputs, limits, defaults and price.\n4. Choose one exact id. In a consultation, name the significant trade-off; in a request to\n   generate, continue the action if no further decision is needed.\n\n## When there is no exact match\n\nA required condition does not become a preference just because the catalog does not support\nit. Offer one useful conditional option: what the user asked for, what a specific model\nsupports and which requirements would have to change. Name every significant concession, not\nonly the first conflict you found. A second option is needed only when it offers a different\nuseful trade-off given the user's priorities.\n\nUse the fields of the current contract, not a closed list of parameters in this skill. Check\nthe combination of mode, inputs and settings: separate maximums do not prove that they are\navailable together. Limits on source material and limits on the result are different.\nMissing information means the capability is unknown, not that it is proven absent.\n\nBefore preparing or proposing a generation, agree with the user on any change to required\nconditions, a replacement of a model they explicitly chose, or dropping materials or required\ncontent. Permission already given for such a trade-off still holds; do not ask again. Combine\nrelated concessions into one concrete question. If the user has not settled a choice between\nconflicting priorities, do not make it for them silently.\n\nFor a chain of steps, check the capabilities of each step and whether their inputs and outputs\nare compatible. Separate the native result from post-processing; report losses, and do not\npromise exact preservation of identity or continuity across a splice. If there is no confirmed\nuseful alternative, explain the obstacle and the smallest change that would let the work\ncontinue. Do not reduce the answer to the word \"impossible\", and do not invent a workaround.\n\nIf the current context is enough, do not reread the catalog or a skill just to explain a\ntrade-off. You need new information only for facts that are missing or have changed.\n\nIf the current catalog names Givon's primary pick for the task, use that recommendation among\nthe compatible models. Do not answer with a list of equal names or with \"it is impossible to\nchoose\" when the request allows a useful primary option. When speed or price is the priority,\nconsider suitable low-cost models, even if they appeared less often in the user's history.\nCheaper does not mean \"draft only\"; do not forbid final use without a concrete limitation.\n\n## Evidence and freshness\n\nDistinguish technical compatibility, Givon's recommendation, a measured result, a provider's\nclaim and creators' practice. Comparisons, benchmarks, popularity and social media examples are\nuseful as dated evidence: take into account the exact version, the task and what exactly was\ncompared. Popularity can explain why a workflow is common, but it does not override\nincompatible inputs or an important limitation.\n\nDo not present a saved review as a continuously updated ranking or trend monitoring. An older\nlist of models does not override the current catalog. If the user asks for a fresh summary,\nuse the available verification tools and name the date; if there are none, separate saved\ninformation from what is not verified right now. An ordinary choice does not require\nresearching the whole internet again: give a primary option based on the relevant evidence you\nhave. A reason belongs in a consultation, not in an obligatory extra reply before every\ngeneration.\n\nCheck a model the user named against its current contract. Do not rank from memory, by provider\nor by keywords, and do not invent a universal \"best\" option.\n"
}