A free GitHub Action that turns every build into EU CRA evidence.
CRA-Check generates a CycloneDX SBOM (via Syft), scans it for known vulnerabilities (via Grype), inspects the repository for security-policy evidence, and emits a machine-readable cra-evidence.json plus a human-readable cra-report.md mapped to EU Cyber Resilience Act requirements (Annex I Part II, Annex VII). A composite Action plus one dependency-free Python script; no server, no account, nothing leaves your CI.
Add a workflow to the repository you want evidence for:
# .github/workflows/cra-check.yml
name: CRA evidence
on:
push:
branches: [main]
pull_request:
jobs:
cra:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: wickitlab/CRA-Check@v1
with:
path: '.'
fail-on: 'off' # set to 'critical' to block releases on critical CVEsSee example-workflow.yml for a complete pipeline. Each run uploads cra-sbom.json, cra-evidence.json, and cra-report.md as build artifacts and posts the report to the GitHub job summary.
| Input | Default | Meaning |
|---|---|---|
path |
. |
Directory to scan; also the repository root for the security-policy evidence checks |
fail-on |
off |
Exit policy: off, high, or critical |
artifact-name |
cra-evidence |
Name of the uploaded evidence artifact |
- A GitHub repository with Actions enabled. That is all: the Action installs pinned Syft/Grype into the runner's temp directory at run time.
- To run the report generator locally: Python 3.10+.
# local run against existing scanner output
python cra_report.py --sbom cra-sbom.json --grype grype.json --repo-dir . \
--out-json cra-evidence.json --out-md cra-report.md| Layer | Technology |
|---|---|
| Action | Composite GitHub Action (action.yml), scanners pinned by version |
| SBOM | Syft (CycloneDX JSON) |
| Vulnerability scan | Grype |
| Report generator | Python 3 (stdlib only, no dependencies) - cra_report.py |
| Tests | pytest (102 tests: 26 for the engine and survey, 76 for the Apify Actor) + CI on every push |
CRA-Check emits 14 gap checks (each pass / warn / fail) across four evidence sources:
| Source | Checks |
|---|---|
| Vulnerabilities | no Critical at release · remediate High+ · no known-exploited (CISA KEV) · no unfixable Critical/High · fix-available triage · scan DB freshness |
| SBOM | SBOM present · product declared as metadata.component · generation provenance (timestamp + tool) · component identifiers (purl/CPE + version) · component license coverage |
| Repository | coordinated vulnerability disclosure policy (SECURITY.md / security.txt) · RFC 9116 security.txt |
| Process | a vulnerability scan was run |
Hard requirements fail (missing SBOM, a Critical at release, a known-exploited CVE, no CVD policy); SBOM-quality and advisory signals warn so they never block a release on their own. Each check cites the CRA Annex I / Article obligation it supports in the report. The severity-based checks use scanner severity as a practical engineering signal toward the CRA's "no known exploitable vulnerabilities" requirement; they are evidence for your assessment, not a legal determination of exploitability.
| Artifact | Purpose |
|---|---|
cra-sbom.json |
Current CycloneDX SBOM (CRA Annex I Part II + Annex VII) |
cra-evidence.json |
Machine-readable CRA gap checks + vulnerability counts |
cra-report.md |
Readable summary, also rendered in the GitHub job summary |
state-of-cra/ aggregates many cra-evidence.json files into one ecosystem readiness report - the share of projects shipping with an SBOM, with unresolved Critical findings, with a disclosure policy, and a per-check pass rate. Run it against the bundled synthetic samples with python state-of-cra/aggregate.py; see state-of-cra/README.md to produce the real survey.
An Apify Actor wrapping this engine is in the repository (apify_actor/, .actor/) and is not published yet. Once it is, it will take an SBOM (CycloneDX or SPDX) by URL or pasted JSON, an optional Grype result and a few product facts, and return the same 14 checks as a dataset item plus cra-evidence.json and cra-report.md, billed per report. With the sealed option it also writes cra-evidence-sealed.json, a signed copy in ProofLog's evidence format that anyone with the publisher's public key can verify later (python -m apify_actor.verify, or ProofLog in .NET); a sealed run is billed as a sealed report instead of a plain one. The report dates follow Article 71 of Regulation (EU) 2024/2847: the Article 14 reporting obligations apply since 11 September 2026 and the Regulation applies in full from 11 December 2027. Until a Store link is added here, treat it as not available. How sealing and verification work is in HOW_IT_WORKS.md; pricing notes for the maintainer are in docs/apify-pricing.md; draft listings for AI-tool directories are in docs/listings/.
- EOL (end-of-life) component detection alongside the CVE checks.
- A
--format htmlreport variant. - Aggregate mode (one report across many repositories).
The dossier generation layer (Annex VII technical documentation, Article 14 incident drafts, continuous CVE watch, tamper-evident evidence ledger) is the commercial CRADesk product line built on top of this Action, together with the ProofLog tamper-evident audit-log SDK and the SecureCheck security-scan workflow.
Not legal advice. Engineering tooling and guidance; it does not by itself guarantee CRA compliance.
This project is licensed under:
- Apache 2.0 - free for any use, commercial or otherwise. See the license for full terms.