| name | license-filter |
| description | Screen EyeDataHub records by source-stated terms and access conditions without making a legal permission determination. |
Skill: source-term and access screening
Use EyeDataHub metadata to narrow records for human review. Never answer that a
dataset is lawful, safe, approved, or commercially usable. EyeDataHub records
source documentation and operational conditions; it does not determine
permission.
Required inputs
- Intended activity and whether it includes redistribution, modification,
model-weight release, commercial use, sensitive-data handling, or clinical
use.
- Modality and task filters.
- Tolerance for registration, authentication, click-through, manual approval,
author contact, payment, and unknown terms.
- Whether counsel, an institutional data office, or an ethics body must review
the source.
Workflow
-
Search independently by modality/task, source-term category, access
friction, availability, and acquisition support.
eyehub search --modality <modality> --task <task> --json
eyehub search --source-terms <category> --json
eyehub search --access <access_friction> --json
-
Inspect each candidate rather than aggregating terms into one verdict.
eyehub show <record> --json
eyehub preflight <record> --json
-
Report the raw source-stated terms, evidence URL, apparent scope, normalized
category, access requirements, route-check date, loader-test scope, and all
unknowns.
-
If terms may apply to code, metadata, a publication, or challenge
participation instead of dataset files, say so. Do not inherit them.
-
Recommend review of the current official source and the appropriate human or
institutional authority. Do not infer a legal result from CC, MIT, Apache,
research-only, unknown, or any normalized group.
Output
Use columns such as record, raw source terms, terms scope, evidence,
access friction, automation/test scope, unknowns, and review needed.
Do not use an “OK for use?” column.
Always state: “This is descriptive source and access metadata, not legal
advice or a permission determination. Review the current official source and
obtain any required institutional or legal approval.”
Refuse
- Converting
standard-no-nc or a legacy commercial-ok alias into permission.
- Treating an unknown as unrestricted.
- Treating a self-service route as reuse authorization.
- Treating a code, paper, website, mirror, or platform licence as the
dataset-file licence.
- Bypassing authentication, click-through, payment, author contact, manual
approval, data-use agreements, geography, or institutional controls.