Earth2Studio Data Source Builder
Connect remote weather data to Earth2Studio with tested source wrappers.
The skill specifies user confirmation for source type and sanity-check plots, and recommends cache cleanup, retries, and excluding secrets from commits. However, confirmations for dependencies, access patterns, and branch creation, commit, push, and PR submission are mostly advisory; credential handling, data-flow disclosure, permission boundaries, rollback, and remote side effects are insufficiently specified, so points are deducted.
The ordered workflow, async retries, concurrency limits, timeouts, exception classes, and try/finally cleanup are reasonably clear, with explicit time and variable validation. However, it depends on reference files not supplied here, lacks an executable test suite within the skill evidence, and static review cannot establish key-path execution, so the score remains conservative.
The audience, argument hint, four source types, explicit non-fit cases, and one-source-per-invocation constraint are clear. Yet it requires a complete Earth2Studio development environment, uv, Git remotes, remote-store access, and possibly credentials; Chinese-language usage and mainland-China network constraints are not addressed, leaving trigger boundaries dependent on agent judgment.
The skill includes name, version, Apache-2.0 license, author, tags, a reference-file table, method ordering, documentation requirements, and CHANGELOG guidance; the README also describes repository maintenance and daily synchronization. However, the benchmark flags the description as overly long, while examples, FAQs, dependency versions, update ownership, and actual changelog details are incomplete, and referenced files are not included in the supplied evidence.
The workflow covers analysis, implementation, tests, registration, documentation, validation, and PR work, so it could support end-to-end datasource integration. However, many steps depend on an external repository, remote data, and unavailable templates, and neither execution, coverage, nor directly usable final outputs are verified in the static evidence; substantial manual correction may remain.
The SKILL.md supplies auditable steps, commands, file locations, and acceptance requirements, and BENCHMARK.md records static-check results. But Tier 3 results are unavailable and there is no committed key-path test or real execution evidence, so the main conclusions cannot be independently reproduced.
- Most confirmation gates are advisory except source type and plots; an agent may create branches, commit, push, or open a PR without explicit approval.
- Private remote stores may require credentials, but injection, log redaction, least privilege, and data-exfiltration boundaries are not specified.
- The referenced implementation-guide.py, testing-guide.py, and validation-guide.md are not included in the supplied material, preventing independent verification of key implementation details.
- The core workflow depends on remote stores, GitHub, and an Earth2Studio environment, with no mainland-China reachability guidance or offline alternative.
What does this skill do, and when should you use it?
This skill guides the development of new Earth2Studio data-source wrappers for remote stores such as S3, GCS, Azure, HTTP, and HuggingFace. It covers source-type selection, dependency analysis, implementation, testing, registration, documentation, validation, and PR preparation. It supports DataSource, ForecastSource, DataFrameSource, and ForecastFrameSource workflows. It is intended for developers adding remote data integrations, not for fetching data through existing sources, model inference, or installation tasks.
Accepts a remote-store URL or description and analyzes its backend, format, authentication, access pattern, temporal and spatial resolution, and variable inventory; checks pyproject.toml and Earth2Studio vocabulary; creates vocabulary mappings, source files, and tests; guides dependency additions, registration, documentation, and CHANGELOG updates; runs pytest, formatting, linting, and optional variable validation; and prepares branch, commit, and PR materials.
- An Earth2Studio developer needs to add a gridded analysis dataset on S3 and wants a DataSource implementation with tests.
- A team needs to integrate a remote forecast dataset containing lead times and wants the appropriate ForecastSource or ForecastFrameSource workflow.
- A developer is integrating data from GCS, Azure, HTTP, or HuggingFace and wants an fsspec-oriented dependency and access analysis.
- A maintainer needs to complete vocabulary mapping, registration, documentation, tests, and CHANGELOG work for a new source in a prescribed order.
What are this skill's strengths and limitations?
- Provides an end-to-end workflow from source selection through testing, validation, and PR preparation.
- Requires source and test files early and specifies engineering patterns for asynchronous data access.
- Covers several remote-store backends and prefers the fsspec ecosystem where applicable.
- Includes concrete guidance for vocabulary mapping, caching, concurrency, retries, and cleanup.
- Requires an Earth2Studio development environment, uv, Git configuration, and access to the target remote store.
- Variable validation and sanity-check plotting require network access and user confirmation.
- The source material provides no concrete implementation, test results, or broad platform testing evidence.
- The repository metadata supplied here lists NOASSERTION; the skill frontmatter says Apache-2.0, while the README describes the collection as dual-licensed Apache-2.0 and CC BY-4.0.
How do you install this skill?
Install the skill from the NVIDIA/skills collection with: npx skills add nvidia/skills --skill earth2studio-create-datasource --yes. The README does not document additional skill-specific installation steps.
How do you use this skill?
With the skill loaded, provide a remote-store URL or description, for example: “Create an Earth2Studio data source for this S3 dataset: https://example.com/data”. If no argument is supplied, the skill asks for a URL, API documentation link, or store description. Each invocation handles one source type.