agentleFS
CycloneDXCycloneDX/cdxgen/.agents/skills/cyclonedx-spec-reviewer/SKILL.md

cyclonedx-spec-reviewer

Skill1.1k starsChanged 14 days agoApache-2.0View on GitHub

Reviews cdxgen changes for CycloneDX schema compliance, semantic field correctness, and unnecessary custom properties across spec versions 1.5 through 2.0, including AI-BOM modeling via formulation, modelCard, pedigree, and evidence. Use when a pull request changes CycloneDX output, schema handling, validators, package-manager integrations, BOM enrichment, or custom properties.

1 other project uses a version of this.

Trending

Quiet this year.about the whole repository

Dec 2025: 0 mentionsJan 2026: 0 mentionsFeb 2026: 0 mentionsMar 2026: 0 mentionsApr 2026: 0 mentionsMay 2026: 0 mentionsJun 2026: 0 mentionsJul 2026: 0 mentionsAug 2026: 0 mentionsSep 2026: 0 mentionsOct 2026: 0 mentionsNov 2026: 0 mentions
Dec 2025Jun 2026Nov 2026
Mentions a month of the CycloneDX/cdxgen repository that our weekly crawl found. There is no adoption data for this file yet.

Most recently taken up by

Who's talking about CycloneDX/cdxgen

The biggest 5 of 41 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 41

Talked about on GitHub and Bluesky (41 times).

On GitHub · 40

35 more on GitHub
  • Other versions1 project uses a version of this.

    Word for word · 1

    1. cdxgen/cdxgen★ 1.1k100% same

    Edited · 0

    None found.

    Wrote their own version · 0

    None found.

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository7 other files it gives its agents.

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. About 0 are common, 9 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when a pull request changes CycloneDX output, schema handling, validators, package-manager integrations, BOM enrichment, or …

    line 8
  2. Distinctive:

    When a change introduces AI-BOM, model inventory, prompt/config discovery, or agent/MCP inventory, explicitly review the standard AI/ML …

    line 22
  3. Distinctive:

    Treat AI-BOM as CycloneDX-first modeling, not as permission to shift standard semantics into custom properties.

    line 32
  4. Distinctive:

    If a custom property is still necessary, require a clear namespace and a narrow purpose.

    line 50
  5. Very specific:

    When a PR adds or changes ecosystem support, check whether emitted components and services are missing expected data that cdxgen can …

    line 78
  6. Distinctive:

    Check whether any new custom property could be replaced by a standard field.

    line 112
  7. Distinctive:

    Check whether any deprecated field is newly introduced or expanded.

    line 113
  8. Very specific:

    Check whether any new package-manager integration omits expected attributes that cdxgen already emits for comparable ecosystems.

    line 114
  9. Distinctive:

    Check whether AI-specific custom properties duplicate standard fields such as model type, task, lineage, datasets, provider, or runtime …

    line 116
  10. Very specific:

    Check spec-version downgrade and upgrade paths, especially 1.5/1.6/1.7 versus 2.0 behavior for AI-BOM structures.

    line 117
  11. Very specific:

    When the intent is to show where a component was found, evidence.occurrences[].location is the more semantically correct field.

    line 140
  12. Distinctive:

    Treat this as a strong signal for removal or redesign rather than a property to preserve.

    line 152
  13. Distinctive:

    When reviewing new integrations, treat missing license, hash, and dependency information as expected completeness checks, not optional …

    line 163
  14. Very specific:

    When a stable Hugging Face, model, or dataset identity is available, prefer durable references (bom-ref, purl, externalReferences, dataset …

    line 169
  15. Very specific:

    If lineage is known only from model-repository metadata, pedigree.notes may be acceptable as supplemental context, but not as the sole …

    line 170

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

CycloneDX specification reviewer

Use this skill when a pull request changes CycloneDX output, schema handling, validators, package-manager integrations, BOM enrichment, or custom properties.

Sources of truth

Review in this order:

  1. Local schemas in data/bom-1.5.schema.json, data/bom-1.6.schema.json, data/bom-1.7.schema.json
  2. Other local schemas when relevant, especially data/cyclonedx-2.0-bundled.schema.json
  3. Official CycloneDX documentation and property-taxonomy guidance

JSON-schema validity is only the starting point. Also review semantic correctness.

AI-BOM focus areas

When a change introduces AI-BOM, model inventory, prompt/config discovery, or agent/MCP inventory, explicitly review the standard AI/ML fields already present in CycloneDX 1.5, 1.6, 1.7, and 2.0.

  • formulation
  • pedigree
  • modelCard
  • data components and componentData
  • services
  • evidence
  • externalReferences

Treat AI-BOM as CycloneDX-first modeling, not as permission to shift standard semantics into custom properties.

What to check

1. Prefer standard fields over custom properties

Flag a change when it introduces or preserves a custom property for data that already fits a standard field such as:

  • supplier, manufacturer, authors, publisher
  • externalReferences
  • evidence.identity
  • evidence.occurrences
  • pedigree
  • hashes
  • licenses
  • scope
  • properties registered in public taxonomy when applicable

If a custom property is still necessary, require a clear namespace and a narrow purpose.

2. Reject ambiguous or risky custom properties

Flag custom properties that are:

  • unnamespaced
  • duplicates of existing CycloneDX fields
  • host-specific or non-reproducible
  • likely to leak local paths, usernames, secrets, or environment-specific details
  • packing structured data into comma-delimited strings when a structured field exists
  • likely to confuse downstream consumers about meaning

3. Check semantic field use

Review whether attributes are used for the right purpose:

  • manufacturer for the entity that created the component
  • supplier for the entity that supplied or distributed it
  • authors for people, not organizations
  • deprecated author only as a compatibility holdover and preferably not in new output
  • publisher only when publication semantics are actually intended
  • externalReferences[].type matches the URL purpose
  • scope is semantically correct for required, optional, excluded, or runtime-only style data
  • bom-ref values are unique and resolvable

4. Check new ecosystem/package-manager integrations

When a PR adds or changes ecosystem support, check whether emitted components and services are missing expected data that cdxgen can usually infer:

  • stable bom-ref
  • type, name, version
  • purl when a native package identity exists
  • dependency edges
  • hashes when resolved artifacts or lockfile integrity data are available
  • licenses when manifest or registry metadata provides them
  • externalReferences for distribution or VCS when directly available
  • evidence fields when provenance or file-location evidence is collected

Flag missing data as a gap when the source material is already available to the integration.

5. Check AI-BOM and AI/ML modeling

Review whether AI-related output uses the standard CycloneDX AI/ML structures correctly:

  • formulation should describe how something was created, trained, assembled, deployed, or otherwise brought into its current form. Do not treat it as a generic dumping ground for unrelated inventory.
  • formulation.workflows, tasks, and steps should represent real process information. If the PR only discovered runtime usage or file-level evidence, verify that this is not being overstated as build/training lineage.
  • machine-learning-model components should prefer modelCard for task, architecture, datasets, and metrics.
  • training or evaluation datasets should prefer modelCard.modelParameters.datasets, ideally via ref to stable type: data components when the BOM already has a stable dataset identity.
  • model lineage such as fine-tunes, distillations, adapters, merges, and quantized derivatives should prefer pedigree.ancestors, pedigree.variants, commits, or patches. Notes may supplement this but should not be the only place where lineage is captured when a structured relation is known.
  • inference endpoints belong in services, not components.
  • prompt files, agent instructions, model config files, and notebooks are usually best represented as file components plus evidence, unless the spec clearly supports a richer standard structure.
  • component.data must only be present for type: data components.
  • component.modelCard should only be present for type: machine-learning-model.
  • service evidence and any 2.0-only structures must be reviewed for spec-version downgrade behavior so that 1.5/1.6/1.7 output does not silently become semantically misleading.

Also review AI-specific custom properties carefully. A namespaced property is still a finding if it duplicates a standard field that now exists in the schema.

Review procedure

  1. Identify the spec version(s) affected by the change.
  2. Compare changed output fields against the local schema definitions.
  3. Check whether any new custom property could be replaced by a standard field.
  4. Check whether any deprecated field is newly introduced or expanded.
  5. Check whether any new package-manager integration omits expected attributes that cdxgen already emits for comparable ecosystems.
  6. For AI-BOM changes, compare every emitted AI field against formulation, modelCard, pedigree, componentData, services, and evidence in the local schemas.
  7. Check whether AI-specific custom properties duplicate standard fields such as model type, task, lineage, datasets, provider, or runtime endpoints.
  8. Check spec-version downgrade and upgrade paths, especially 1.5/1.6/1.7 versus 2.0 behavior for AI-BOM structures.
  9. Produce findings in four groups: spec violation, semantic misuse, unnecessary customization, and expected data missing.

Expected review output

For every finding, include:

  • exact field or property name
  • why it is a problem
  • preferred CycloneDX field or modeling approach
  • schema/spec basis
  • whether it is a repo-wide legacy pattern or introduced by the PR

Repo calibration findings

These repo-specific findings were identified while calibrating this skill and should be treated as known review heuristics.

  • Self-generated BOMs for this repository currently expose three unnamespaced custom property names: internal:SrcFile, internal:ImportedModules, and internal:LocalNodeModulesPath.

Iteration 1: legacy internal:SrcFile property

  • Current cdxgen output emits unnamespaced internal:SrcFile properties.
  • In self-generated BOMs this often duplicates evidence.identity[].methods[].value.
  • When the intent is to show where a component was found, evidence.occurrences[].location is the more semantically correct field.

Iteration 2: legacy internal:ImportedModules property

  • Current cdxgen output emits unnamespaced internal:ImportedModules.
  • The value is packed as CSV-like text and mixes module names with symbol-like data.
  • This is hard for downstream tooling to interpret and should be reviewed against evidence.occurrences plus symbol, or replaced with a clearly namespaced property if no standard field fits.

Iteration 3: legacy internal:LocalNodeModulesPath property

  • Current cdxgen output emits unnamespaced internal:LocalNodeModulesPath.
  • It can contain absolute host paths, which are environment-specific and potentially sensitive.
  • Treat this as a strong signal for removal or redesign rather than a property to preserve.

Iteration 4: deprecated and incomplete producer metadata

  • Current output may rely on deprecated metadata.component.author instead of authors or manufacturer.
  • Root BOM metadata commonly has metadata.authors but not metadata.supplier.
  • For published artifacts, review whether supplier and contact/manufacturer data should be emitted instead of or in addition to author-style fields.

Iteration 5: completeness gaps for emitted components

  • Self-validation of a generated BOM for this repository still shows components missing license data and at least one component missing hashes/dependency linkage.
  • When reviewing new integrations, treat missing license, hash, and dependency information as expected completeness checks, not optional polish, if the upstream ecosystem exposes that data.

Iteration 6: AI-BOM review heuristics

  • cdx:ai:* properties are namespaced and therefore better than unnamespaced properties, but they should still be challenged when they duplicate standard CycloneDX AI/ML fields.
  • In particular, review whether task, lineage, datasets, provider identity, runtime identity, and variant classification should move to modelCard, pedigree, externalReferences, services, or evidence.
  • When a stable Hugging Face, model, or dataset identity is available, prefer durable references (bom-ref, purl, externalReferences, dataset refs) over free-text notes.
  • If lineage is known only from model-repository metadata, pedigree.notes may be acceptable as supplemental context, but not as the sole representation when the relation itself is structured and reproducible.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.