CRA Index / Open dataset
CRA Index open dataset
The nine yes/no security signals behind the CRA Index for 303 popular packages (200 npm, 103 PyPI), free to download as CSV or JSON under CC BY 4.0. Version 2026.11.16, data read 2026-10-26 to 2026-11-16.
Download
- cra-index-packages.csv (one row per package)
- cra-index-packages.json (same rows with the signals grouped, plus metadata)
- Data dictionary and as JSON
- Changelog ยท Licence and attribution
The nine signals in plain words
Each signal is true (present in public data), false (not present) or empty (the data is not published, so we do not know).
| Column | What it asks | True | Not known |
|---|---|---|---|
no_known_vulns | No known vulnerabilities. deps.dev lists no advisory for the default version. | 100 % | 0 |
not_deprecated | Not deprecated. The registry does not mark the version as deprecated. | 100 % | 0 |
actively_maintained | Actively maintained. The Scorecard Maintained check is 7 or higher (recent commits, issues handled). | 44 % | 43 |
security_policy | Security policy. The Scorecard finds a security policy (for example SECURITY.md) that scores 6 or higher. | 66 % | 43 |
licence_declared | Licence declared. The registry names a licence. | 98 % | 0 |
code_review | Code review before merge. The Scorecard Code-Review check is 7 or higher. | 24 % | 45 |
ci_hygiene | CI workflow hygiene. No dangerous workflow patterns and sensible token permissions in the build workflows. | 36 % | 70 |
signed_or_provenance | Signed releases or provenance. Build provenance or attestations are published, or releases are signed. | 41 % | 21 |
security_testing | Automated security testing. The repository uses static analysis (SAST) or fuzzing. | 33 % | 43 |
True is the share among packages where the signal could be read, in this version. Exact thresholds are in the data dictionary.
Where the data comes from and how often it refreshes
Package facts, licences, deprecation, advisories and provenance come from deps.dev (its generated data is CC BY 4.0; advisories come from OSV). Repository signals come from the OpenSSF Scorecard, whose results are CDLA Permissive 2.0. We read them with a deterministic script, apply the thresholds and publish only the resulting yes/no values, the package name and version, and the dates. Raw Scorecard scores, advisory identifiers, repository addresses and star counts are not part of the release. The dataset is rebuilt weekly together with the index; each release gets a version (the date of the latest read) and a changelog entry.
What this does not show
- It is not a compliance assessment. Nothing here says whether a package, a project or a product meets the Cyber Resilience Act.
- False is not bad. A project may keep a policy, review process or signing elsewhere, or use a tool the Scorecard does not detect.
- No judgement of people. There are no names, contacts or maintainer data, and no ranking of projects by quality.
- Not the full ecosystem. 303 popular packages, picked from public popularity lists. Small, new and private packages are absent, so shares do not describe npm or PyPI as a whole.
- Snapshot, not a feed. Advisories and Scorecard results change; check the date in each row. Scorecard results are only available for packages whose source is on GitHub.
- No dependency data. Signals describe the package itself, not what it depends on.
How to cite
The licence is CC BY 4.0: use it freely, including commercially, and credit Ambolt and the upstream sources named in the licence file. Corrections and questions: hello@ambolt.dev.
Browse the same data as pages: npm, PyPI.
Information generated from your inputs, with article references to Regulation (EU) 2024/2847. It is not legal advice and does not replace your own assessment; check the references against the Official Journal.