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

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

@@ -0,0 +1,60 @@
+---
+name: hf-paper-search
+description: "Search for ML research papers on Hugging Face and arXiv, read methodology sections, trace citation graphs, and find linked datasets and models."
+disable-model-invocation: false
+---
+
+# hf-paper-search — Paper Research
+
+## Purpose
+
+Find and analyze ML research papers to extract training recipes, datasets, hyperparameters, and known failure modes before implementing.
+
+## Tools
+
+- `paper_search`: Search HF papers by query. Good for quick discovery.
+- For deeper operations (details, section reading, citation graphs, snippet search, recommendations, linked resources), use the paper research script:
+
+```bash
+python skills/ml-intern-harness/scripts/papers.py <operation> [args]
+```
+
+## Paper Script Operations
+
+| Operation | Description |
+|---|---|
+| `search` | Search papers with Semantic Scholar filters (date, citations, categories) |
+| `trending` | Get daily trending papers |
+| `paper_details` | Get abstract, AI summary, GitHub link |
+| `read_paper` | Read arXiv/ar5iv HTML sections |
+| `citation_graph` | Find references or citations |
+| `snippet_search` | Semantic search across paper passages |
+| `recommend` | Find similar papers |
+| `find_datasets` | Find HF datasets linked to a paper |
+| `find_models` | Find HF models linked to a paper |
+| `find_collections` | Find HF collections linked to a paper |
+| `find_all_resources` | All linked resources at once |
+
+## Workflow
+
+1. Use `paper_search` for quick discovery or `papers.py search` for filtered Semantic Scholar search.
+2. Read methodology and results sections with `papers.py read_paper --arxiv-id <id> --section 3`.
+3. Trace citation graphs with `papers.py citation_graph --arxiv-id <id> --direction citations`.
+4. Extract concrete recipes: dataset, preprocessing, method, hyperparameters, model, metric, result.
+5. Find linked HF datasets/models/collections with `papers.py find_all_resources`.
+6. Inspect promising datasets with `inspect_dataset.py`.
+7. Read current HF docs and GitHub examples before implementing.
+
+## Example
+
+```
+_paper_search(query="direct preference optimization")
+python skills/ml-intern-harness/scripts/papers.py read_paper --arxiv-id 2305.18290 --section 3
+python skills/ml-intern-harness/scripts/papers.py citation_graph --arxiv-id 2305.18290 --direction citations --limit 10
+python skills/ml-intern-harness/scripts/papers.py find_all_resources --arxiv-id 2305.18290
+```
+
+## Research Standard
+
+For paper-backed tasks, attribute important choices to a source:
+- "Dataset X + method Y + model Z produced metric M on benchmark B" (source: arXiv:2305.18290, Section 4.2)."