One SBOM standard for a polyglot company
By SBOMBox SBOM experts · · 3 min readShort answer: agree on one contract, not one tool. Pick CycloneDX 1.6 JSON, generate from the lockfile in CI for every release, leave dev dependencies out, name the root component after the product and version, require purls and hashes, check quality before sharing, sign it, and keep it for the support period. Then give every repository the same entry point that enforces those rules.
The problem: same word, different documents
Ask five teams for their SBOM and you will get five documents that disagree on format, version, scope and naming. The Node team's includes test frameworks; the Python team's has no hashes; the Java team's is SPDX; the platform team's describes the container, not the app. Each is defensible on its own. Together they are useless to the security, legal and compliance people who need to answer one question across all of them: are we affected?
An eight-rule standard
- One format: CycloneDX 1.6 JSON internally. Convert to SPDX only when someone outside asks. See CycloneDX vs SPDX.
- One source of truth: the committed lockfile, generated in CI, for every release.
- One scope: runtime dependencies in the release SBOM.
- One naming rule: root component type
application, name is the product, version is the release tag. - Identifiers: a purl on every component, hashes wherever the lockfile has them.
- A quality floor: schema-valid, non-empty, and above an agreed score.
- Integrity: signed or attested against the artifact digest.
- Retention: immutable, per product and version, for the support period. See how often to update an SBOM.
One entry point
The rules only stick if teams do not have to remember them. Put a single script in a shared CI template that chooses the right generator for the lockfile it finds, applies the flags, and refuses to produce an empty file. When a better generator arrives for an ecosystem, you change one file. The unify guide has a tested version that uses npm's, pnpm's and uv's built-in commands and syft for everything else.
./sbom.sh services/payments payments.cdx.json
cyclonedx validate --input-file payments.cdx.json --fail-on-errorsMerge per product, not per repository
Customers and regulators ask about products. Generate one SBOM per service, then merge them hierarchically into one SBOM per product release with the CycloneDX CLI, so each service stays visible as its own branch of the tree. Keep the container image's SBOM as another branch rather than mixing its OS packages into the application's list.
Keep watching after release
Most vulnerabilities that affect a release are published after it ships. Store every release's SBOM and re-scan all of them on a schedule, and record each 'not affected' decision as VEX so nobody triages it twice.
Where a platform helps
Every step above can be done with open-source tools and some CI templates. The cost is maintaining them. SBOMBox runs the same lockfile-first generator choice as sbombox generate, scores and checks compliance on every upload, records verified provenance, stores each release immutably and re-checks it nightly. You can adopt it one step at a time and keep your own generators: sbombox push takes any CycloneDX or SPDX file. The end-to-end workflow shows how the pieces fit.
Frequently asked questions
Do all teams need the same generator?
No. They need output that meets the same rules. Use the best generator for each ecosystem behind one entry point.
Should dev dependencies ever be in an SBOM?
In a separate build SBOM, if you want to track build-time supply-chain risk. Not in the release SBOM you give customers, where they look like shipped code.