# AwesomeCUA

> A source-backed computer-use agent resource catalog. Distinguish curated entries from metadata-only bibliography records.

Resource counts are not independent project counts. Primary-source retrieval does not establish reproducibility. Code, data, weights, and document licenses are separate.

## Start here

- [Catalog](https://awesomecua.github.io/catalog/): resource categories
- [Research](https://awesomecua.github.io/papers/): reviewed papers and surveys
- [中文论文库](https://awesomecua.github.io/zh/papers/): 按年份浏览论文与综述
- [Guides](https://awesomecua.github.io/guides/): definitions and practical reading guides
- [Editorial policy](https://awesomecua.github.io/about/): review states and limits
- [Coverage](https://awesomecua.github.io/coverage/): source populations and known gaps
- [Data downloads](https://awesomecua.github.io/data/): structured catalog with provenance

## Machine-readable access

- [Full catalog JSONL](https://awesomecua.github.io/downloads/catalog.jsonl)
- [Curated annotations in Markdown](https://awesomecua.github.io/llms-full.txt)
- [Research Atom feed](https://awesomecua.github.io/feed.xml)

Resource HTML pages have a companion `index.md` at the same path. This file is an optional navigation aid, not an indexing requirement or a request to rank this site.

## Compare source-reviewed specifications

- [Benchmark protocols](https://awesomecua.github.io/compare/benchmarks/)
- [Dataset profiles](https://awesomecua.github.io/compare/datasets/)
- [Benchmark profiles JSON](https://awesomecua.github.io/downloads/benchmarks-profiles.json)
- [Dataset profiles JSON](https://awesomecua.github.io/downloads/datasets-profiles.json)

## Agent and RAG access

```sh
git clone https://github.com/reacher-z/Awesome-CUA.git
```

- [Agent instructions](https://awesomecua.github.io/AGENTS.md)
- [Human and agent guide](https://awesomecua.github.io/agents/)
- [中文指南](https://awesomecua.github.io/zh/agents/)
- [Reviewed retrieval documents](https://awesomecua.github.io/downloads/rag-documents.jsonl)
- [All retrieval documents](https://awesomecua.github.io/downloads/rag-all-records.jsonl)
- [Export manifest](https://awesomecua.github.io/downloads/rag-manifest.json)

## Individual records and citations

Each active resource page serves its complete record at `record.json`. Scope-excluded resources return a JSON object with `status: excluded`, not an active record.

- [Reviewed paper BibTeX](https://awesomecua.github.io/downloads/citations.bib)
- [Reviewed paper CSL JSON](https://awesomecua.github.io/downloads/citations.csl.json)
- [Citation manifest](https://awesomecua.github.io/downloads/citations-manifest.json)

## Publication editions

Default arXiv citations retain the recorded preprint. Publication editions use their own title/authors/date and do not create extra catalog papers. Date components keep their recorded precision.

- [Reviewed publication BibTeX](https://awesomecua.github.io/downloads/publication-citations.bib)
- [Reviewed publication CSL JSON](https://awesomecua.github.io/downloads/publication-citations.csl.json)
- [Publication observations](https://awesomecua.github.io/downloads/publications.jsonl)
- [Publication observation schema](https://awesomecua.github.io/schemas/publication-observation.schema.json)

Resources with publication metadata also serve `publication-citation.bib` and `publication-citation.csl.json` next to their retained-source citation files.

## Topic collections

- [Topic directory](https://awesomecua.github.io/topics/)
- [中文主题目录](https://awesomecua.github.io/zh/topics/)
- [Topic manifest](https://awesomecua.github.io/downloads/topics.json): exact reviewed resource IDs, counts, subset URLs and SHA-256 fingerprints. Topic memberships overlap; counts must not be summed.

The first page of each topic has an index.md listing all of its resources. Each /downloads/topics/<topic>/ directory serves records.jsonl, citations.bib and citations.csl.json. Pass records.jsonl to export_rag.py --input to build a topic-specific retrieval corpus.

## Repository identity

- [Verified former repository names](https://awesomecua.github.io/downloads/repository-aliases.json): paired official API observations with the same GitHub or Hugging Face repository ID. Use the canonical returned ID for retrieval upserts; aliases add no resources. Record fields and RAG documents retain former names and their alias-scoped evidence.
- [Repository identity schema](https://awesomecua.github.io/schemas/repository-alias.schema.json)

## Download inventory

- [Bilingual data hub](https://awesomecua.github.io/data/)
- [中文数据中心](https://awesomecua.github.io/zh/data/)
- [Download manifest](https://awesomecua.github.io/downloads/manifest.json): file paths, exact byte sizes, SHA-256 fingerprints, counts and units. Compare catalog_sha256 with the RAG, citation and topic manifests before combining artifacts.
- [Download manifest schema](https://awesomecua.github.io/schemas/download-manifest.schema.json)

## Versioned data

- [Tagged data releases](https://github.com/reacher-z/Awesome-CUA/releases): complete data bundles containing catalog, RAG, citations and topic subsets. Verify SHA256SUMS and retain release-manifest.json, source_commit and catalog_sha256. Repository CLI commands require a source clone at the recorded commit.

## English research survey

- [Computer-use agent survey](https://awesomecua.github.io/survey/): working manuscript with a separately frozen research inventory; not an additional catalog record.
- [Survey bibliography](https://awesomecua.github.io/survey/references.bib)
- [Survey references JSON](https://awesomecua.github.io/survey/references.json)
- [Source reading register](https://awesomecua.github.io/survey/evidence/reading-index.json)
- [Survey file manifest](https://awesomecua.github.io/survey/build-manifest.json)