Upload plugins/mlintern/skills/hf-model-search/SKILL.md with huggingface_hub

plugins/mlintern/skills/hf-model-search/SKILL.md

@@ -0,0 +1,48 @@
+---
+name: hf-model-search
+description: "Search Hugging Face Hub for models, verify model metadata, architecture, tokenizer, and task fit before using a model."
+disable-model-invocation: false
+---
+
+# hf-model-search — Hugging Face Model Discovery
+
+## Purpose
+
+Find and validate Hugging Face models before using them in training, inference, or evaluation.
+
+## Tools
+
+Use the following tools to discover and inspect models:
+
+- `model_search`: Search HF Hub models by task, author, tags, or query string.
+- `hub_repo_details`: Get detailed metadata for a model repo: README, tags, downloads, likes, library, task, and config files.
+
+## Workflow
+
+1. Search for candidate models with `model_search`.
+2. Inspect promising candidates with `hub_repo_details` (set `repo_type="model"`, `include_readme=true`).
+3. Verify:
+   - Model architecture matches the intended task.
+   - Tokenizer/config files are present.
+   - License is compatible.
+   - Gating/token requirements are understood.
+   - Downloads/likes suggest community validation.
+4. If the task requires a specific model variant (quantized, fine-tuned, GGUF), note the specific branch or commit.
+
+## Example
+
+```
+_model_search(query="sentence embedding", task="sentence-similarity", sort="downloads")
+_hub_repo_details(repo_ids=["sentence-transformers/all-MiniLM-L6-v2"], repo_type="model", include_readme=true)
+```
+
+## Validation Checklist
+
+Before depending on a model:
+- [ ] Repo exists and is not a redirect.
+- [ ] `config.json` or equivalent config is present.
+- [ ] Tokenizer files (`tokenizer_config.json`, vocab) are present.
+- [ ] Task tags match the intended use.
+- [ ] Library tag (transformers, sentence-transformers, etc.) is compatible.
+- [ ] License is acceptable.
+- [ ] Gating is handled (token set, request submitted).