Changelog
Release history from GitHub
sbom-tools v0.3.0
Highlights
The CBOM fidelity release. v0.2.0 checked whether the tool told the truth; this release asks the same of cryptography inventories, prompted largely by people running real CBOMs through it. A CycloneDX 1.7 IPsec CBOM did not load at all. Crypto fields renamed between CycloneDX 1.6 and 1.7 were read in only one spelling, so an elliptic-curve key declared the 1.6 way, or a certificate linked the 1.7 way, slipped past the CNSA 2.0 and PQC gates. And a sparse IKE CBOM drew 31 findings for missing package versions, suppliers, licenses and hashes on its crypto assets, while the quality score undercounted its real problems — HMAC-SHA1, 3DES and a 1024-bit DH group.
All of those are fixed, and fixtures in both spec versions pin the behaviour.
Alongside the CBOM work: convert --to normalized exposes the full normalized model on the command line, byte-identical to the C ABI payload; CycloneDX 1.6+ manufacturer and authors now count as producer evidence across nine compliance standards; Yocto SPDX 3 documents parse; and a first, unsigned pipeline receipt contract lands under sbom-tools verify.
Please read Upgrade notes below — this release changes several compliance verdicts in both directions, the standard the CBOM quality profile reports, machine-readable output, and the Rust API.
What's New
Cryptography (CBOM)
- CycloneDX 1.7 IKEv2 transforms. Both the 1.6 bom-ref string and the 1.7 object parse, in JSON and XML. CNSA 2.0 and PQC
PROTO-002judge each transform by its algorithm reference, IANA name or key-exchange group, including the ML-KEM groups 35–37; a transform that cannot be identified reportsPROTO-UNKNOWNrather than passing. (#380) - Crypto fields are read in both their 1.6 and 1.7 spellings — the curve, a certificate's signature algorithm and public key, a key material's algorithm, and protocol refs (
cryptoRefArray/relatedCryptographicAssets). A 1.7 certificate that relates several algorithms has every candidate judged, failing closed. (#378) - Crypto assets are not packages. Version, supplier, license, hash and identifier checks no longer apply to
cryptographic-assetcomponents under NTIA, FDA, Standard, Comprehensive, CISA 2026, PCI DSS and FSCT, or in the quality metrics, andquality --profile cbomembeds NIST PQC compliance instead of Comprehensive, which ran no crypto checks. (#382) - One algorithm classifier for
qualityandvalidate, so the two agree:HMAC_SHA1counts as weak, and IKE groups (MODPnnnn,ECPnnn,curve25519,curve448) are classified. The CycloneDX 1.7 protocol typesike,quic,wpa,sstp,eap-aka,eap-aka-prime,prinsand5g-akaare recognised rather than falling through toOther. (#382)
CLI
convert --to normalizedprints the full normalized model — crypto, ML and dataset detail included — byte-identical to the payload the C ABI and language bindings return, which was previously reachable only by linking the library. (#367)- Pipeline receipts.
verify receipt,verify receipt-generateandverify receipt-aggregatevalidate, generate and aggregatepipeline-shard-receipt/v1documents against three versioned JSON Schemas. Receipts are unsigned and non-promotable diagnostics; seedocs/pipeline-receipts.md. (#353)
Compliance and AI-BOM
- Component
manufacturerandauthors(CycloneDX 1.6+) are parsed, emitted, and accepted as producer evidence by NTIA, FDA, CRA, CISA 2026, EO 14028, SSDF, both BSI profiles and PCI DSS. (#380) - AI-003 accepts nested dataset evidence — a
containschild carrying a CycloneDXdata[]block — alongsidemodelParameters.datasets; a bare nestedtype: datacomponent still does not count. A third-party AI-BOM corpus andscripts/test-aibom.shpin the behaviour. (#361)
Documentation
- JSON output contract in the README: which payloads are contractual, what is test-pinned, and that
view -o jsonis a curated projection without crypto, ML or dataset detail. (#364)
Bug Fixes
- CycloneDX 1.7 IPsec CBOMs failed to load — an IKEv2 transform written as a 1.7 object stopped the parser with exit
3. (#380, #379) - The CNSA 2.0 and PQC gates failed open across spec versions. A 1.6 EC key identified only by
curvereported an UNKNOWN Warning instead of a quantum-vulnerable Error, and 1.7 certificates linked throughrelatedCryptographicAssetswere skipped byCERT-001. (#378, #374) - CBOMs were graded on package fields — 31 package-only findings on a sparse IKE CBOM under
quality --profile cbom, 19 undervalidate --standard ntia— while the quality score undercounted weak crypto and could not classifymodp1024. (#382, #370) - The protocol warning claimed "documents no version" for protocols that declared one. It now names the fields that are actually missing. (#382, #370)
- Components that named a
manufacturerorauthorsfailed supplier rules as though they were anonymous. (#380) - Yocto SPDX 3 documents failed to parse because
suppliedBywas given as a single string rather than an array. (#349, #348) - The TUI Violation Detail overlay clipped its content at common terminal heights, cutting off remediation text and the close hint, and the CRA supply-chain rule's remediation gave dependency advice instead of naming the supplier fields. (#355, #347)
- Security. RUSTSEC-2026-0258 (
h2), RUSTSEC-2026-0221 (event-listener), RUSTSEC-2026-0253 (lru, use-after-free), RUSTSEC-2026-0285 (rustls) and RUSTSEC-2026-0338 through -0342 (quinn-proto) are cleared, with no advisory suppressions. (#352, #369, #388)
Upgrade notes
These behaviors changed in this release. CHANGELOG.md carries the full list with details.
- Some CBOM gates are stricter. A 1.6 EC asset identified only by
curvenow failsALG-006, 1.7 certificates and protocols linked throughrelatedCryptographicAssetsare now evaluated, and 1.7 IPsec CBOMs that failed to parse are now judged. A passingvalidaterun can now exit1. - Other gates are looser. Crypto assets no longer get package findings under NTIA, FDA, Standard, Comprehensive, CISA 2026, PCI DSS or FSCT, and a named
manufacturerorauthorsnow satisfies supplier rules. A failing run can now exit0. quality --profile cbomreports NIST PQC compliance instead of Comprehensive, and CBOM scores move in both directions — the #370 fixture goes from 42.1 to 37.2.- JSON gains fields, additively:
cryptography_metrics.protocols_with_algorithm_refsin quality output;related_crypto_assets, componentmanufacturer, and IKEv2authandtransformsin the normalized payload. - Library API:
ComponentgainsmanufacturerandIkev2TransformTypesgainsauthandtransforms; code that builds them with struct literals must add the fields.ProtocolTypegains eight variants.
Acknowledgments
Thanks to @lvlrSajjad (Sadjad (Mo) Asadi) for the issues behind the curve fix and the normalized-JSON work (#362, #366, #374), to @paul007ex for the CBOM evaluation in #370, to @iwanicki92 for reporting #347 and both reporting and fixing #348 (#349), and to @MChorfa (Mohamed Chorfa) for the pipeline-receipt contract (#350, #353) and the CI baseline, benchmark and security-gate repairs (#351, #369).
Install: cargo install sbom-tools
Homebrew: brew install sbom-tool/tap/sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.2.0...v0.3.0
sbom-tools v0.2.0
Highlights
The correctness release. Where v0.1.22 added an AI-BOM domain, this release goes back over what was already there and checks whether it told the truth. Systematic audits of the diff engine, the parser layer, enrichment, quality scoring, the compliance engine, the CLI surface and the TUI turned up defects that shared a shape: the tool reporting something other than what it had computed. A demo SBOM shipped in this repository reported "no vulnerabilities" while its own Source view displayed a CVSS 9.8 entry. Fleet deviations rendered as 10000.0%. A document containing no cryptography scored a perfect 100 for quantum readiness. A keyboard overlay advertised keys that did nothing.
All of those are fixed, and most are now pinned by tests that fail if the behaviour regresses.
Alongside the audit work: three new compliance standards bring validate --standard to 16 — CISA's 2026 Minimum Elements (published July 2026, successor to the NTIA 2021 elements), PCI DSS v4.0.1 Requirement 6.3.2, and CISA Framing Software Component Transparency 3rd ed. — plus CDXA attestation ingestion, so a signed machine-readable declaration can satisfy an SSDF practice or a CRA conformity route that previously accepted only self-declaration. A new quality -o sbomqs-json emits the sbomqs 0–10 score model for side-by-side comparison. The project also gains its first CHANGELOG.md and a TUI keyboard reference.
Please read Upgrade notes below — this release changes machine-readable output, the Rust API, several compliance verdicts, and the TUI keymap.
What's New
Compliance
- CISA 2026 Minimum Elements —
validate --standard cisa-2026(aliasescisa,minimum-elements-2026), 17 rules against the final v2.1 document published 2026-07-29. Deliberately stricter than the NTIA profile where the document is: a tool-only creator list does not satisfy SBOM Author, and a silently absent license fails where an explicit no-assertion passes. (#333) - PCI DSS 6.3.2 —
validate --standard pci-dss, 10 rules treating the SBOM as the required inventory of bespoke, custom and third-party components. A passing verdict is evidence the inventory exists and is usable, not a PCI compliance certification, and thresholds chosen by this tool rather than the standard are labelled as such in the message. (#333) - CISA FSCT 3rd ed. —
validate --standard fsct, 27 rules across the document's three maturity tiers, mapped Minimum Expected → Error, Recommended Practice → Warning, Aspirational → Info. Checks are format-gated where the evidence cannot exist rather than failing a document for its format. (#333) - CDXA attestation evidence. CycloneDX 1.6
declarationsare parsed into a normalized model, and a declaration covering an SSDF practice, a CRA conformity route, an EUCC certificate reference or EO 14028 provenance now satisfies that rule at structural evidence level. Every pre-existing satisfaction path remains valid as a self-declared fallback, so documents without declarations are unaffected. Verification is structural only — signature presence is recorded, never cryptographically verified. (#333, #334) - Pinned evaluation clock.
--as-offixes the date deadline-sensitive checks evaluate against, so CI runs are reproducible instead of drifting with the wall clock. (#320) - Registry-driven rule identity across JSON, SARIF, the quality CLI and OSCAL, plus tri-state applicability so "not applicable" is distinct from "compliant". (#320)
- Per-rule test matrix. Every rule in the FDA, SSDF, EO 14028, Standard and Comprehensive profiles now has a firing and a silent fixture, with exhaustiveness enforced in both directions so a new check site forces a matrix entry. (#333)
Interoperability
quality -o sbomqs-jsonemits the sbomqs v2.0.11score --jsonschema, and the plain-text quality summary gains a 0–10 category table. Formulas were read from that release's sources rather than its documentation, including the behaviours a reimplementation would otherwise "correct"; categories that cannot be computed are emitted asignoredwith a stated reason instead of approximated. (#333)
Parsers
- SPDX 3.0.1 JSON-LD documents parse, including canonical
@graphforms. (#311) - SPDX 2.x data-loss gaps closed —
documentDescribes, file entries, andLicenseRefresolution. (#315) - CycloneDX correctness — spec-valid documents that hard-failed now parse, nested assemblies are inventory, and serialization round-trips. (#314)
- purl-shaped references resolve.
dependencies[].refandvulnerabilities[].affects[].refgiven as purls now match components, with bom-refs still winning on collision. Documents written this way previously lost dependency edges and vulnerabilities silently. (#333) - Document revision tracking. The CycloneDX top-level
versioncounter is captured, diffed, and round-tripped throughconvert, which had hardcoded"version": 1. (#333, #334)
TUI
- Four phases of work land here: data surfacing and a theme backbone (#309), per-tab depth with detail panels and ML/VEX/match transparency (#310), a multi-mode overhaul with shared builders (#312), and unified search behind a single shortcuts surface (#313).
- A UX audit of every mode against a 269-frame corpus produced 203 findings; the 30 that survived adversarial verification are fixed, along with the tail. Empty states no longer celebrate, every advertised key exists,
ycopies the row you actually selected, and the Quality, Compliance and AI-Readiness tabs no longer collapse to empty borders at 80×24 — the app's own advertised minimum size. (#333, #334) docs/TUI_SHORTCUTS.mddocuments every binding across all three TUIs, each traced to its handler with the guard that makes a key mean different things on different tabs. (#334)
Documentation
CHANGELOG.md, which the project had never had. Every claim in it and in the refreshedREADME/docswas verified against the built binary and then re-checked by independent adversarial review. (#334)
Bug Fixes
- Vulnerabilities and dependency edges were silently dropped whenever a CycloneDX document referenced components by purl — including this repository's own demo fixtures, where the Vulnerabilities tab reported a clean bill while the Source tab displayed a CVSS 9.8 CVE. (#333)
- Fleet deviation was double-scaled, rendering as
10000.0%and saturating every gauge and severity band. It is now a 0–1 fraction, and the three similarity scales are pinned by tests. (#333) - A version bump counted as both an addition and a removal in multi-SBOM analysis, so "Inconsistent" totals exceeded the number of distinct components and Timeline listed every upgrade twice. (#333)
- Documents with no cryptography scored 100 for quantum readiness. The score is now absent rather than vacuously perfect, and the CBOM scorer redistributes the weight. (#333)
- The diff engine, parser, enrichment, quality and CLI audits closed the solver/matching/incremental-cache defects (#299), 29 untrusted-input parser findings (#300), VEX correctness and network trust surface (#303), compliance false-passes and score math (#304), and the CLI exit-code contract and fail-closed gates (#322).
- Compliance engine overhaul — correct verdicts, verified citations, one output contract, and the removal of vacuous CRA passes. (#320)
diffandviewaccepted-o sbomqs-jsonand silently emitted ordinary JSON under a format the caller never asked for. Both now reject it with a pointer toquality. (#334)--no-colordid not reach the TUI, which consulted only theNO_COLORenvironment variable, and the side-by-side reporter ignored both. An emptyNO_COLOR=now counts as unset everywhere, per the convention. (#333, #334)- The shortcuts overlay contradicted itself — its global rows were gated by mode but not by tab, so one frame could assert both
l Color legendandh/l Collapse/expandon a tab wherelswitches standards. (#334) - Post-merge end-to-end testing closed four further bugs in emitted VEX, LicenseRef surfacing, SPDX creator/file emission, and SARIF family fallback. (#321)
Upgrade notes
These behaviors changed in this release. CHANGELOG.md carries the full list with details.
- Deviation is a 0–1 fraction (
diff-multionly).summary.deviation_scoresandsummary.max_deviationare now0.0…1.0; they were double-scaled before.timeline -o jsonhas nosummaryobject and is unaffected. - Multi-SBOM component identity is the version-stripped logical purl (
pkg:npm/express, notpkg:npm/express@4.18.2), so a version bump is one component with a version spread. diff -o jsongainscomponent_typeon component changes anddoc_versionmetadata entries. Additive, but strict consumers must accept them. Change totals move: the demo fixture pair goes from 15 to 18.diff-multi,timelineandmatrixaccept only-o auto|tui|json. An unsupported value is a usage error (exit2) at parse time rather than a runtime failure (exit3).- Library API:
CryptographyMetrics::{quantum_readiness_score, pqc_readiness_score}returnOption<f32>;run_tui/run_view_tuitake ano_colorparameter. validate --standard ntiaand--standard fdaare engine-backed and stricter — SBOMs that passed before can now fail.- CRA phases re-anchored to Reg. (EU) 2024/2847 — Phase 1 is Art. 14 reporting from 11 Sep 2026, Phase 2 is full application from 11 Dec 2027. Article and Annex citations were corrected throughout.
validate --standard bsiis rebased on BSI TR-03183-2 v2.1.0, and several SARIF/JSONrule_ids were renamed to registry-driven identity.- Exit codes fail closed. Operational errors exit
3, command-line parse errors exit2, and a nonzero exit is a gate verdict only when the expected report was produced. - The TUI keymap changed. Digits always jump tabs; Components' quick filters moved to a
Qpicker, Side-by-Side filters toA/r/m/x, the KEV filter toK, vulnerability group jump to}/{, Dependencies fold tox/X, the cycles toggle toC, and the threshold overlay — previously unreachable — tot.docs/TUI_SHORTCUTS.mdis the full reference;?in the app is context-aware.
Install: cargo install sbom-tools
Homebrew: brew install sbom-tool/tap/sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.22...v0.2.0
sbom-tools v0.1.22
Highlights
The AI BOM release. sbom-tools now treats AI systems as a first-class domain across the whole pipeline: it parses CycloneDX ML-BOMs and SPDX 3.0 AI/Dataset profiles, scores AI-readiness (AI-001…AI-010), machine-checks the EU AI Act (Annex IV) and the G7/BSI "SBOM for AI — Minimum Elements" standards, diffs model and dataset changes semantically, verifies model-weight integrity, enriches Hugging Face components, and renders AI-BOMs as a dedicated profile in the TUI (Models / Datasets / AI-Readiness tabs).
Alongside the AI work this release lands a large wave of correctness and security fixes — including a hostile-SBOM stack-overflow, empty OSV severity data, and broken CycloneDX XML parsing — a new convert command (CycloneDX ↔ SPDX), KEV/EPSS vulnerability enrichment with an offline/air-gapped mode, and broad CLI/TUI alignment.
Please read Upgrade notes below — this release changes several user-visible behaviors (scoring engine 2.1, stricter license-policy gating, logs moved to stderr, official EPSS endpoint).
What's New
AI / ML Bill of Materials
- AI-readiness scoring profile. A dedicated
quality --profile ai-readiness(and TUI AI-BOM profile) grades ten AI-specific checks (AI-001…AI-010): model-card, architecture, training datasets, quantitative analysis, fairness, ethical considerations, use-cases, limitations, energy, and model-weight hash integrity. (#205, #239, #246) - CycloneDX ML-BOM + SPDX 3.0 AI/Dataset parsing. Spec-compliant CycloneDX
modelCardparsing (typed fairness/quantitative-analysis/considerations) and SPDX 3.0 AI + Dataset profile metadata including thetrainedOnrelationship. (#204, #208, #239) - Semantic ML model & dataset diffing. Model and dataset changes are diffed field-by-field — quantization, architecture, training-dataset add/remove, dataset sensitivity — with high diff-costs on provenance-loss and PII-escalation signals, instead of opaque blobs. (#244)
- Model-weight integrity & Hugging Face enrichment. New AI-010 weight-hash check, a
verify model-weights --model-dirsubcommand (direct + HF cache layouts), and a Hugging Face enrichment source that mapssiblings[].lfs.sha256→ component hashes,pipeline_tag→ task, and routespkg:huggingfacemodels into the vulnerability stack. (#246, #253) SBOM-AIBOM-*SARIF rule family for AI-readiness findings, and TUI rendering of ML model / dataset metadata in the component-detail panel. (#207, #206)
Compliance
- EU AI Act (Annex IV) readiness compliance level with
SBOM-AIACT-*SARIF rules, N/A-gated for non-AI SBOMs and severity-escalated by the CRA high-risk-AI flag. (#245) - G7 / BSI "SBOM for AI — Minimum Elements" readiness — a new
validate --standard bsi-aiprofile that scores an AI-BOM element-by-element across the seven BSI clusters (Metadata, System-Level, Models, Datasets, Infrastructure, Security) withSBOM-BSIAI-*rules. (#255) - Compliance rule registry. Every violation now carries a stable
rule_id; SARIF rule IDs and standard references come from a single registry instead of fragile message-text matching. (#231, #233)
Cross-format conversion
convertcommand. Re-emit any parsed SBOM to CycloneDX 1.7 or SPDX 2.3 with a fidelity report listing synthesized/dropped fields, built on an opt-in source-field preservation slot. (#241, #243, #247)
Vulnerability enrichment
- CISA KEV and EPSS enrichment. New
--kev/--epssflags surface Known-Exploited-Vulnerabilities flags and EPSS exploit-probability on vulnerabilities, with--fail-on-kevgating, KEV/EPSS columns in reports, and a watch-mode "entered KEV" alert. (#235, #237) - Offline / air-gapped mode + cache management. A global
--offline(andSBOM_TOOLS_OFFLINE) serves enrichment purely from cache, plus a newcache status / warm / clear / export / importsubcommand for sneakernet transfer. (#238) - Unified enrichment platform. All sources (OSV, EOL, VEX, KEV, staleness, EPSS, Hugging Face) now share one
EnrichmentSourcetrait, an atomic-write cache with schema versioning, and a single retry/User-Agent HTTP client. (#230)
CLI & TUI
- Global
--configfile is now honored by every command (it was previously read only byconfig show), with CLI-flag > file > default precedence. (#227) - Stdin input (
-as a path) fordiff/quality/validate/query/vex, enablingsyft … | sbom-tools quality -pipelines. (#224) - Typed exit codes & NDJSON output. Per-command exit codes,
clapvalue-enums (typos fail at parse), and a wired-o ndjsonformat. (#226) - Document-metadata diffing.
diffnow reports author, tool, timestamp, spec-version, signature and serial-number changes. (#254) - AI-BOM is first-class in the TUI — detected as its own profile with AI-readiness scoring and dedicated Models / Datasets / AI-Readiness tabs, plus EPSS/KEV badges, a KEV filter, richer component detail, and a profile-aware help overlay. (#258, #257, #259, #261, #260, #223)
Bug Fixes
- Hostile-SBOM stack-overflow fixed. Cycle detection is now iterative (Tarjan SCC); a deeply-nested dependency chain previously aborted the process (and any FFI host) via stack overflow. (#215)
- OSV enrichment now carries severity / CVSS / description. Results from
/v1/querybatch(id-only) are hydrated via/v1/vulns/{id}; previously enriched vulnerabilities had empty severity and were cached empty for 24h. (#219) - CycloneDX XML parsing fixed for spec-conformant documents (components, dependencies, vulnerabilities, license choice, xmlns spec version). (#214)
- License policy correctness.
fail_on_conflictnow actually fails, concluded-only licenses are evaluated against the deny list, and deny-patterns match per-operand in SPDX expressions (deny "GPL-*"now catchesMIT OR GPL-3.0-only). (#212) - Per-component license changes are populated in diffs (the field was always empty), and the Licenses TUI tab no longer falsely reports "no changes". (#211, #259)
- Watch mode no longer fires false "resolved vulnerability" alerts on file changes (re-parsed SBOMs are now enriched before diffing). (#221)
- Incremental diff cache no longer splices sections from an unrelated cached pair, and diff errors propagate instead of becoming "no changes". (#217)
- Deterministic diffs. Fixed-seed MinHash and stable solver ordering make diff output reproducible above the LSH threshold (CI
--fail-on-*gating). (#218) - Machine output is parseable again. Logs are written to stderr so
-o json/-o sarif/-o ndjsonon stdout is valid when piped or redirected (previously twoINFOlines preceded the JSON, breaking SARIF upload andjq). (#256) - Multi-SBOM commands honor their flags.
diff-multi/timeline/matrixnow apply--graph-*/filter/rules and can reach exit code 4, and reject unsupported-ovalues. (#225) - Output hardening. CSV formula-injection guard, terminal control-char sanitization, and a TUI panic hook that restores the terminal. (#213)
- Enrichment entry-point consistency.
enrich/watch/querynow route through the unified orchestrator so--kev/--epss/--huggingfacetake effect everywhere,query --offlineno longer makes network calls, and the EPSS default endpoint is the official FIRST host with a response-size bound. (#249, #248, #247) - FFI binding parity. The
AiReadinessprofile is exposed in the Go / Swift wrappers with a CI drift check, and thesbom-tools-fficrate's version/cargo publishblockers are fixed. (#216, #234)
Performance
- Sparse assignment matching. The dense O(n²) Hungarian path (≈200 MB / effective hang on cross-format diffs) is replaced by a sparse solver over the candidate edge list. (#232)
- In-place enrichment mutation removes a 2× component-map copy per pass, and report-stage compliance is computed lazily per reporter. (#236)
- TUI dependency tab caches its graph and uses an iterative longest-path, eliminating a per-frame rebuild and an exponential depth calculation. (#223)
Internal & Quality
- Compliance engine split into a
StandardCheckerper-standard module layout behind a registry, and the enrichment platform unified behind one trait. (#233, #230) - cli / tui feature gates so the FFI staticlib and fuzz targets no longer link ratatui/clap/rustls (FFI dependency tree 281 → 104 crates). (#228)
- Test infrastructure. A ratatui
TestBackendsnapshot harness for the TUI, golden snapshots for all report formats (plus streaming-JSON validity fixes), graph-shaped hot-path benches, and diff/scoring property tests. (#229, #240, #242, #218)
Infrastructure
- Dependency bumps:
ratatui0.30.0 → 0.30.1,chrono0.4.44 → 0.4.45,regex1.12.3 → 1.12.4,reqwest0.13.3 → 0.13.4,serde_json1.0.149 → 1.0.150,libfuzzer-sys0.4.12 → 0.4.13, plus dagger-SDK-only bumps. New runtime dependency:flate2(gzip, enrichment feature) for the EPSS bulk feed. (#251, #252, #195, #196, #222, #248) - GitHub Actions bumps:
codeql-action→ 4.36.1,cargo-deny-action→ 2.0.20,actions/checkout→ 6.0.3; SHA-pin hygiene + expanded Dependabot coverage; fuzz CI now installscargo-fuzzwith the nightly toolchain. (#193, #194, #199, #200, #202) - Total tests: 1560 (1014 lib + 546 integration), up from 1214 in v0.1.21.
- 0 clippy warnings (default + all-features) on Rust 1.88; 0 production
unwrap().
Upgrade notes
These behaviors changed in this release:
- Scoring engine 2.0 → 2.1. Cycle detection now counts strongly-connected components (actual cycles) rather than DFS back-edges, so quality scores for SBOMs with cyclic dependencies may change. (#215)
semantic_scorenow counts per-component license transitions, so diffs that change a component's license score slightly lower than before. (#211)- License-policy gating is stricter.
license-checknow fails on license conflicts and concluded-only denied licenses that previously passed silently — CI gates may newly fail (correctly). (#212) - Logs are on stderr. Anything parsing tool stdout for log lines must now read stderr; machine-readable report output on stdout is clean. (#256)
- EPSS default endpoint moved to the official FIRST host
epss.empiricalsecurity.com(gzip). Override with--epss-url/SBOM_TOOLS_EPSS_URL. (#248) - Library API:
IncrementalDiffEngine::diffandMultiDiffEngine::{diff_multi,timeline,matrix}now returnResult. (#217)
Acknowledgments
The AI-BOM direction in this release was driven by @MChorfa (Mohamed Chorfa), who proposed the original ML-BOM and AI-readiness work and filed the issues this release implements: AI-BOM model-card scoring (#184), SBOM-AIBOM SARIF rules (#185), SPDX 3.0 AI-profile parsing (#186), and TUI ML/dataset display (#187). Thanks also to @jkowalleck (Jan Kowalleck) for the CycloneDX Tool Center invitation (#176) behind the new submission guide (#177), and continued thanks to @cmyank0 and @VincentR-OCD whose issue reports keep sbom-tools' diff and enrichment output honest.
This release also builds on the public security and standards work behind the G7/BSI "SBOM for AI" minimum elements, the EU AI Act, CISA KEV, and FIRST EPSS.
Install: cargo install sbom-tools
Homebrew: brew install sbom-tool/tap/sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.21...v0.1.22
sbom-tools v0.1.21
Highlights
Focused bug-fix release. view -o json now emits the full enriched vulnerability detail (not just counts), diff semantic similarity is correctly bounded to 0–100 (previously could exceed 100% on large diffs), and CRA compliance sections in Markdown/HTML diff reports collapse to a compact summary so reports stay readable on SBOMs with hundreds of components.
Bug Fixes
view -o jsonnow emits enriched vulnerability detail. The JSON view previously dropped enriched OSV/KEV data and emitted only a per-component count, even when--enrich-vulnswas set. Each component now serializes the full vulnerability list (id, source, severity, CVSS,fixed_version, CWEs, KEV info, VEX status, description, dates), and the document gains a top-levelvulnerabilities[]array flattened across all components. Each component and flattened entry is also tagged withdependency_kind(primary/direct/transitive). (#179, fixes #178)diffsemantic similarity bounded to 0–100.DiffResult::semantic_scoreis normalized against an SBOM-size-derived upper bound and clamped to[0, 100], fixing reports of values like 916% on heavy-churn diffs. Identical SBOMs (matching content hash) now correctly report 100.0. (#188, fixes #180)- Compact CRA compliance section in human-readable diff reports. Markdown and HTML diff reports previously emitted a full violation table per CRA finding, ballooning to 14k+ lines on real SBOMs. The diff path now emits a compact "N findings across M groups" summary with a pointer to
-o json/-o sariffor full structured detail. Single-SBOMviewreports and JSON/SARIF outputs are unchanged. (#189, fixes #181)
Infrastructure
- Dependency bumps:
quick-xml0.39.3 → 0.40.1,clap_complete4.6.3 → 4.6.5,tokio1.52.2 → 1.52.3 (dagger SDK only). - GitHub Actions bumps:
codeql-action4.35.3 → 4.35.5,cargo-deny-action2.0.17 → 2.0.18,crates-io-auth-actionSHA-pin refresh. - Total tests: 1214 (787 lib + 427 integration), up from 1196 in v0.1.20.
- 0 clippy warnings (default + all-features) on Rust 1.88; 0 production
unwrap().
Acknowledgments
Thanks to @MChorfa (Mohamed Chorfa) for contributing the fixes behind all three user-facing improvements in this release (#179, #188, #189), and to @cmyank0 and @VincentR-OCD for the bug reports that drove them.
Install: cargo install sbom-tools
Homebrew: brew install sbom-tool/tap/sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.20...v0.1.21
sbom-tools v0.1.20
Highlights
sbom-tools v0.1.20 ships end-to-end EU Cyber Resilience Act (CRA) readiness: prEN 40000-1-3 normative requirement IDs, CRA product-class & conformity-route severity calibration, Article 24 open-source software steward profile, BSI TR-03183-2 §5/§6 checks, CSAF v2.0 ingest and emit (ISO/IEC 20153:2025), and a cra-docs CLI that prefills Annex V documentation directly from the SBOM + sidecar. The release also adds CRA standards-drift detection to watch, with alerts routed through the existing AlertSink so they show up alongside SBOM file changes and vulnerability discoveries — no new daemon, no new TUI tab.
Combined with the CBOM scoring engine shipped in v0.1.19 and the existing NIS2 / GDPR / AI Act / RED overlap surfaces, sbom-tools is now the most complete open-source CRA conformity-assessment toolkit covering Articles 13/14, Article 24, and Annex I–VIII obligations.
What's New
CRA Compliance Engine (Articles 13/14, Article 24, Annexes I–VIII)
- 5 new
ComplianceLevelvariants (9 → 14):Cnsa2(NSA CNSA 2.0),NistPqc(FIPS 203/204/205 + IR 8547),BsiTr03183_2(German national CRA-aligned guideline),CraOssSteward(Article 24 lighter profile),EuccSubstantial(Reg. (EU) 2024/482 reference profile) --cra-product-class {Default|ImportantClass1|ImportantClass2|Critical}onvalidate,quality,view: drives Annex III/IV severity calibration, conformity-route gating (Module A / B+C / H / EUCC), vendor-hash thresholds (50/80/80/100%), and severity ramp for EOL components, DoC presence, Article 14 PSIRT readiness, and EUCC referenceConformityAssessmentSummary+ConformityEvidenceonComplianceResult: per-route checklist (DoC + Module-specific evidence + PSIRT). Surfaced in markdown / HTML reports and TUI compliance view asAnnex VIII: <route> (<class>) N/M evidenceis_oss_stewardsidecar profile: SBOM + vuln-handling + CVD policy floor enforced; manufacturer email, EU DoC, module attestation, hardware [PRE-8-RQ-02], and vendor-hash carry-through skipped. Aliases:--standard oss-steward/cra-oss/cra-art24/art24
CSAF v2.0 (ISO/IEC 20153:2025)
- CSAF ingest in
src/enrichment/vex/csaf.rs: auto-detect priority CSAF → CycloneDX VEX → OpenVEX. Mapsproduct_status(known_affected/known_not_affected/fixed/first_fixed/under_investigation/recommended/last_affected) toVexState. Resolvesproduct_id→ PURL viafull_product_names+ recursive branches + relationships; CVE →ids[].textfallback for vuln ID - CSAF emit via new
vex export --format csafCLI: closes the round-trip — SBOM → CSAF → re-ingest preserves identical VEX states. Maps{Affected|NotAffected|Fixed|UnderInvestigation}→ CSAF product status keys; CVE/GHSA dual-path on vulnerability IDs
Commands & CLI
cra-docs: new subcommand that generateseu-declaration-of-conformity.md(Annex V DoC),technical-documentation.md(Annex V tech-doc summary, with NIS2 / GDPR / AI Act / RED overlap section when respective sidecar flags are set), andvulnerability-handling-policy.md(Annex I Part II) — prefilled from sidecar + SBOM +ComplianceChecker. Auto-discovery extended to multi-extension stems (app.cdx.jsonresolvesapp.cra.json)cra-standards-watch: new subcommand that prints a curated 8-artefact catalogue (prEN 40000-1-3, prEN 40000-1-2, BSI TR-03183-2, CSAF v2.0, ENISA SBOM Guidance, EUCC, STAN4CRA, ETSI EN 303 6xx) with last-known versions and--check-onlineHTTP-status probevex export --format csaf: CSAF v2.0 advisory emitter (see above)watch --cra-standards(+--cra-standards-interval=24h,--cra-standards-timeout=10s): periodically probe the CRA standards catalogue from the watch loop and surface drift through existing alert sinks
Watch & AlertSink
AlertSink::on_cra_standard(default no-op trait method): carriesCraStandardEventwithInitialBaseline/StatusChangedkinds — keeps the trait extensible without churning third-party sinks- Stdout sink: timestamped human-readable line per drift (
[12:34:56] cra-standard BSI-TR-03183-2: status drift 200 OK -> 503 Service Unavailable) - NDJSON sink: structured
{"type":"cra_standard","kind":"status_changed",...}records for SIEM/jq pipelines - Webhook sink: deliberately stays at the trait default — drift signals are low-value to spam to webhooks
Reporting & SARIF
SARIF helpUripopulated on rules, plusresults.standardHelpUris, drawn from canonical EUR-Lex / BSI / NIST / OASIS URLs. NewStandardKind::canonical_help_uri()andSarifRuleWithUriserialization wrapperdocs/CRA_COMPLIANCE.mdreverse-mapping table covering CRA Articles 13/14, all Annexes, Article 24, BSI TR-03183-2, CSAF v2.0, the P3.2 calibration table, a CLI cheat sheet, a sidecar example, and a standards bibliography. Linked from README
TUI
- New policy presets:
EuccSubstantial,CraOssSteward(alongside the existingCra/Ntia/Fda/NistSsdf/Eo14028) - Conformity-assessment summary rendered on the compliance view per active product class
- Sidecar plumbing so CRA-specific scoring + class-aware severities reach the TUI without round-tripping through CLI
Compliance Levels — Now 14
Up from 9 in v0.1.19: Minimum, Standard, NtiaMinimum, CraPhase1, CraPhase2, FdaMedicalDevice, NistSsdf, Eo14028, Cnsa2, NistPqc, BsiTr03183_2, CraOssSteward, EuccSubstantial, Comprehensive. ComplianceLevel::all() auto-reflects every variant into the TUI compliance tabs.
Bug Fixes
cra-docssection numbering: Annex V tech-doc previously emitted duplicate5.xheadings when the NIS2 / GDPR / AI Act / RED overlap section was active — now correctly renumbers- TUI sidecar plumbing: compliance view no longer dropped CRA-specific severities when the sidecar was loaded from disk (audit-fix landed alongside P5 polish)
- clippy 1.88 cleanup:
uninlined_format_argsandderivable_implsin CSAF/CRA paths
Infrastructure
- 24 new CRA-product-class tests (
tests/cra_product_class_tests.rs) - 12 new OSS-steward tests (
tests/cra_oss_steward_tests.rs) - 9 new conformity-assessment-summary tests + 8 new SARIF-helpUri tests + 9 new CSAF tests + 7 cra-docs integration + golden CRA fixtures (compliant + 2 non-compliant variants)
- Total tests: 1196 (787 lib + 409 integration), up from ~802 in v0.1.19
- Dependabot bumps:
clap_complete4.6.2 → 4.6.3,reqwest0.13.2 → 0.13.3,quick-xml0.39.2 → 0.39.3,dagger-sdk0.20.4 → 0.20.8 (rust-sdk),tokio1.51.1 → 1.52.2 (rust-sdk),indexmap2.13.1 → 2.14.0,clap4.6.0 → 4.6.1 - Security fixes:
rand,rustls-webpki(PR #145, PR #156) - GitHub Actions bumps:
crates-io-auth-action,codeql-action4.35.2 → 4.35.3,cosign-installer4.1.1 → 4.1.2,cargo-deny-action2.0.15 → 2.0.17,upload-artifact7.0.0 → 7.0.1 - 0 clippy warnings (default + all-features) on Rust 1.88; 0 production
unwrap()
Install: cargo install sbom-tools
Homebrew: brew install sbom-tool/tap/sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.19...v0.1.20
Highlights
sbom-tools now includes a full CBOM (Cryptographic Bill of Materials) quality scoring engine — the first open-source CLI tool to grade cryptographic inventory health alongside traditional SBOM quality. CBOMs receive crypto-specific scores across 8 categories: algorithm strength, PQC readiness, OID coverage, crypto completeness, key/certificate lifecycle, cross-reference resolution, provenance, and licenses. Hard caps enforce grade ceilings for broken cryptography (D max), compromised keys (C max), and zero post-quantum migration (C max).
Combined with the CBOM architecture shipped in v0.1.18 (CycloneDX 1.6/1.7 cryptoProperties parsing, CNSA 2.0 + NIST PQC compliance, 5 dedicated TUI tabs), sbom-tools now provides end-to-end CBOM lifecycle support — from parsing to scoring to interactive analysis to compliance reporting.
What's New
CBOM Quality Scoring Engine
ScoringProfile::Cbom: New 8-category scoring profile with crypto-tuned weights — Algorithm Strength (0.22), PQC Readiness (0.15), Crypto Completeness (0.15), OID Coverage (0.15), Crypto Lifecycle (0.13), Cross-References (0.10), Provenance (0.08), Licenses (0.02)- 12 granular metrics:
algorithms_with_oid,algorithms_with_family,algorithms_with_primitive,algorithms_with_security_level,certs_with_signature_algo_ref,keys_with_algorithm_ref,protocols_with_cipher_suites,keys_with_state,keys_with_protection,keys_with_lifecycle_dates,certs_with_validity_dates - 6 per-category scoring methods: Each computes a 0-100 score from real field coverage (no proxies) — completeness measures family/primitive/level coverage, identifiers measure OID presence, lifecycle blends key state/protection/dates with certificate validity
- Hard caps for critical crypto issues: Broken algorithms (MD5, DES, RC4) → D max; compromised keys → C max; zero quantum-safe algorithms → C max
- Auto-detection:
BomProfile::detect()triggers CBOM scoring when ≥3 crypto components make up >50% of the BOM — works in bothviewanddiffmodes
Commands & CLI
quality --profile cbom: Evaluate CBOM quality from the command line with crypto-specific grades and recommendations- Diff mode awareness: Both old and new SBOMs auto-detect CBOM profile, each scored with the appropriate engine
TUI
- Dynamic quality chart: Bar labels switch to
Crpt/OIDs/Algo/Refs/Life/PQC/Prov/Licin CBOM mode; header shows crypto category names for strongest/weakest - Crypto Inventory panel: Asset counts, quantum readiness %, and warning flags for weak algorithms, compromised keys, and expired certificates
Pkey toggle: Runtime SBOM/CBOM switch — re-scores with the selected profile- Algorithm sorting (
skey): Cycle Name → Family → Quantum Level → Strength (weak-first triage) - Certificate sort by urgency: Auto-sorted by days remaining (expired first)
- Algorithm detail: Now shows Padding, Execution Environment, Implementation Platform
Bug Fixes
- CBOM tab selection: Arrow keys now correctly move selection on all crypto tabs (Algorithms, Certificates, Keys, Protocols) — detail panel was stuck on the first item due to reading a legacy shared index instead of per-tab selection
- Diff mode scoring: Both SBOMs now auto-detect profile instead of hardcoding Standard
Infrastructure
- 15 new CBOM unit tests (scoring methods, hard caps, edge cases)
- Bump indexmap 2.13.1, semver 1.0.28, dagger-sdk 0.20.4
- Bump GitHub Actions: crates-io-auth-action, setup-go 6.4.0
Install: cargo install sbom-tools
Homebrew: brew install sbom-tool/tap/sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.18...v0.1.19
Highlights
This release delivers a comprehensive diff engine upgrade with enrichment support for all multi-SBOM commands, a complete TUI overhaul across all 10 tabs, and introduces Go and Swift bindings for the core SBOM operations.
What's New
Diff Engine
- Multi-SBOM enrichment:
diff-multi,timeline, andmatrixcommands now support--enrich-vulns,--enrich-eol, and--vex(previously onlydiffhad enrichment) - Pre-release version handling: New
SemverPartsstruct properly distinguishes1.2.3-alphafrom1.2.3with pre-release penalty in matching scores - Quality delta tracking:
QualityDeltain diff results shows per-category score changes, grade transitions, regressions, and improvements across all 8 quality categories - VEX state transitions:
VexStatusChangedetects when persistent vulnerabilities change VEX state between SBOMs (e.g., NotAffected → Affected) - Match quality metrics:
MatchMetricsrecords exact/fuzzy/rule match counts, unmatched components, and confidence scores - True incremental diff: Section-selective recomputation skips unchanged sections when cache detects partial changes (30–70% speedup for partial SBOM updates)
TUI — Summary Tab
- Redesigned layout: Merged Risk Assessment + Key Findings into a single compact card with natural-language insights
- Key Findings section: Prioritized action items — critical vulns, major version bumps, license conflicts, quality regressions, added/removed summaries
- SBOM Comparison table: Old vs New delta table showing component, dependency, license, and vulnerability counts
- All Changes list: Full scrollable list sorted by semver impact (MAJOR first) with inline version transitions
- "Not enriched" warning: Vulnerability card shows hint when enrichment wasn't used
TUI — Quality Tab
- Compact gauge header: Score transition in title with text gauge bars (replaces heavy bordered gauges)
- Category bar chart: Sorted by score with colored bars, warning badges (⚠ low, ✗ zero), and delta arrows
- Linked recommendations: Each action shows its target category; Enter navigates to the related tab
TUI — Compliance Tab
- Compact header: 3-line card replacing 7-line side-by-side gauges with New/Resolved counts
- Violation grouping:
gkey groups violations by component with expand/collapse - Short tab labels: Min/Std/NTIA/CRA-1/CRA-2/FDA/SSDF/EO14028/Full — all fit on screen
- Rich empty states: When Resolved/New views are empty, shows top issues to fix and category breakdowns
- View cycle: Changed from
Tab(conflicted with global tab switching) tov
TUI — Diff (Side-by-Side) Tab
- Unified upgrade view: New
Unifiedmode (akey) matches removed+added by name to show version upgrades - Semver classification: MAJOR/minor/patch/DOWNGRADE badges sorted by impact
- Full-width single panel: Replaces split view for clearer upgrade visualization
TUI — Dependencies Tab
- Sort order works (
skey): Roots sort by Name, Depth, Vulnerability count, or Dependent count — previously displayed but never applied - Transitive toggle works (
tkey): Filters to direct deps only (depth 0–1) when off — previously displayed but had no effect - Relationship badges: Child nodes show
[dev]/[build]/[test]/[opt]for non-default dependency types; detail panel shows Relationship and Scope fields - Child count breakdown: Root nodes show color-coded
(+4 -3)instead of plain(7) - Vulnerability section in detail panel: Up to 5 CVEs with severity coloring, status (introduced/resolved/persistent), and KEV badge
- Detail panel scrolling:
Ctrl+d/Ctrl+uscrolls long content; resets on navigation - Compact 2-line header: Removed redundant key-hint line (saves 1 row), color-only toggle states,
+N/-Mchange summary,Expanded: N/Mprogress, full words for alerts - Depth badges: D0/D1/D2 indicators on tree nodes showing direct vs transitive
- Enriched detail panel: Licenses, hashes, supplier, dependency/dependent lists with shared renderer
TUI — Source Tab
- Change gutter:
+/-/~column next to line numbers colored by change type - Change summary bar: Horizontal minimap showing change distribution across the file
- Difftastic-style rendering: Changed lines bold+colored, unchanged lines keep syntax highlighting
- Bold line numbers: Line numbers colored on changed lines for visual emphasis
- Inline version diff: Modified components show
lodash 4.17.20 → 4.17.21when collapsed - Collapse unchanged (
u): Folds unchanged regions into··· N unchanged items ··· - Aligned diff panels (
a): Gap·····placeholders for added/removed components
TUI — Cross-Tab
- Consistent detail panels: Shared
render_component_info_lines()for Components and Dependencies - Cross-tab navigation: License→Components, Violation→Component, Component→Dependencies with breadcrumbs
- Regex search: Ctrl+R toggles substring/regex mode with error feedback
- Footer hints: Added to Compliance, Source, Vulnerabilities, Dependencies tabs
- SBOM version in header: Shows
name@versionfrom primary component
CLI
- Enrichment arg deduplication:
SharedEnrichmentArgswith#[command(flatten)]across all 8 enrichment-capable commands - Bug fixes:
--no-colorand--export-templateforwarded to multi-SBOM commands;--fail-on-vulnexit codes for timeline/matrix - Output flag alignment:
license-checkandverify audit-hashesmigrated from-fto-owith backwards-compat aliases - Feature gaps filled:
--fail-on-change,--exclude-vex-resolved,--fail-on-vex-gap,--graph-max-depthfor multi-SBOM commands - Comprehensive help: 24 top-level examples in 9 categories, per-subcommand examples on 13 of 15 commands
Go and Swift Bindings (MVP)
- C ABI layer: Format detection, parsing, diffing, and quality scoring via JSON-based ABI
- Go wrapper:
sbomtools.ParsePath(),Diff(),Score()with typed payloads and dedup helpers - Swift wrapper:
SbomTools.parsePath(),diff(),score()with Codable payloads and dedup helpers - ABI contract snapshots: Schema conformance tests ensure stability across releases
Thanks to @MChorfa for contributing the Go and Swift bindings!
Bug Fixes
- Dependencies crash: Footer copy-preview panicked on multi-byte
→in dependency edge text (&text[..27]on non-char-boundary) — replaced with MSRV-compatiblefloor_char_boundary()across 10 slicing sites - Dependencies search OOB:
next_match/prev_matchcould index pastvisible_nodeswhenselectedwas desynchronized — added bounds clamping - Dependencies detail panel: Fixed "No node selected" (visible_nodes not populated in prepare_render); fixed duplicate "Name:" line showing canonical ID
- Quality navigation: Fixed arrow keys not working (total_recommendations never set)
- Compliance Tab key conflict: View cycle changed from Tab to
v(was captured by global tab switching) - Matching asymmetry: Proptest found alias-lookup asymmetry for short names — excluded from test generator
Infrastructure
- Bump sha2 0.10→0.11, proptest 1.10→1.11, clap_mangen 0.2→0.3, tempfile 3.25→3.27
- Bump GitHub Actions: codeql-action 4.35.1, setup-go 6.3.0, cosign-installer 4.1.1
- Dagger Rust SDK runner for bindings CI/CD
Install: cargo install sbom-tools
Homebrew: brew install sbom-tool/tap/sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.17...v0.1.18
Highlights
This release hardens the release pipeline against tag poisoning attacks (like the recent Trivy incident), significantly improves the Source tab experience in the TUI, and fixes a severity badge bug that could underreport vulnerability severity.
What's New
TUI
- Source tab overhaul: Fix flickering by hoisting state mutations out of the render path
- Raw view breadcrumb: Show JSON path context at the top of the raw view
- Compact tree mode (
Ckey): Narrower connectors save ~30% horizontal space at depth 4+ - Sticky scope header: Opening bracket line pins to the top when scrolled off-screen
- Active scope indent guide: Single dashed
┆in accent color replaces cluttered multi-level guides - Dependency bom-ref resolution: UUIDs in Source tab dependencies section now show component names (e.g.,
gcc@12.2.0 → 3 deps) - Dependency context in SBOM Map panel: Right panel shows resolved component name, dep count, vulnerabilities, and license instead of generic SBOM stats
- Smart path truncation: Long file path labels show
…/basenameinstead of truncating mid-path - Cleaner link indicators: Remove noisy
⏎symbols and skip redundant labels on expandable objects - Severity badge spacing: Add space between badge and count in dependency tree (
[L] 6vs[L]6)
Security
- Tag poisoning hardening: Ancestry checks on
build-binariesandupdate-homebrewjobs reject commits not onmain - Immutable Homebrew URLs: Formula source pinned to commit SHA instead of mutable tag ref
- Signed release manifests: Sigstore-signed
release-manifest.sha256for release asset tamper detection - Daily tag integrity workflow: Verifies all release tags are on
mainand release asset checksums match signed manifests - Tag protection ruleset: Tags are protected against force-push and deletion
- Homebrew tap protection: Branch and tag rulesets on
sbom-tool/homebrew-tap - Artifact verification docs: SECURITY.md now documents
gh attestation verify,cosign verify-blob, and SHA256 verification
Bug Fixes
- Severity badge underreporting:
max_severity_from_vulns()now derives severity from CVSS base scores when the categoricalseverityfield isNone, fixing cases where badges showed[L]despite high CVSS scores
Infrastructure
- Collapse 47 nested
ifstatements into let-chains (0 clippy warnings) - Fix RUSTSEC-2026-0044, RUSTSEC-2026-0048, RUSTSEC-2026-0049 (
aws-lc-sys0.39.0,rustls-webpki0.103.10) - Bump clap 4.6.0, clap_complete 4.6.0, clap_mangen 0.2.33, tracing-subscriber 0.3.23
- Bump GitHub Actions: codeql-action 4.33.0, rust-cache 2.9.1, crates-io-auth-action
Install: cargo install sbom-tools
Homebrew: brew install sbom-tool/tap/sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.16...v0.1.17
Highlights
Major format expansion with CycloneDX 1.7 and SPDX 3.0 support, three new CLI commands (enrich, verify, license-check), VEX integration for vulnerability exception tracking, and a comprehensive TUI overhaul with ViewState architecture, RenderContext, and dozens of UX improvements. Releases now include pre-built binaries for 5 platforms with Sigstore signing, build attestations, and Homebrew tap distribution.
What's New
Format Support
- CycloneDX 1.7 — full parser, model, quality scoring, and compliance support
- SPDX 3.0 — JSON-LD parser, Security profile (VEX/CVSS), Snippet, Annotation, Lifecycle, multi-document, streaming optimization
- CycloneDX 1.7 enrichment — structured citations support
New Commands
enrich— enrich SBOMs with vulnerability data (OSV/KEV), EOL detection, and VEX statementsverify— file hash verification (SHA-256/512) and component hash auditinglicense-check— license policy engine with allow/deny/review lists and propagation analysisvex— apply, filter, and report on VEX statements; CycloneDX VEX document parsing; coverage summary with--fail-on-vex-gap
TUI
- ViewState architecture — single source of truth for all 10 tab states (Phases 1-6)
- RenderContext — read-only render abstraction eliminating render-time mutation
- Document Info panel — metadata overview in header area
- Export dialog — export current view to file
- Component overview — quick summary of component counts and types
- Quality tab — improved chart rendering, integrity/provenance panels
- Compliance tab — severity filter (
fkey), better gap visualization - Dependencies tab — graph cache improvements, visual polish
- Source tab — cross-tab navigation, SBOM Map alignment, raw mode parity, focus panel fix
- UI consistency — footer hints, keyboard shortcut alignment, visual refinements across all tabs
Distribution
- Homebrew —
brew install sbom-tool/tap/sbom-tools - Pre-built binaries — 5 platforms (linux x86_64/aarch64, macOS x86_64/aarch64, Windows x86_64)
- Sigstore signing — keyless signatures on all release archives
- GitHub build attestations — per-binary provenance
- Dual-format SBOMs — CycloneDX 1.6 + SPDX 2.3 included in release assets
- GitHub Action —
sbom-tool/sbom-tools-actionfor CI/CD integration
VEX Integration
- VEX subcommand with apply/status/filter operations
- CycloneDX VEX document parsing
- Multiple VEX responses per vulnerability
- Persistent gap tracking across diff operations
- Coverage summary and
--fail-on-vex-gapCI gate
Bug Fixes
- Fixed 7 high-severity dependency vulnerabilities (aws-lc-sys, quinn-proto)
- Fixed focus panel not resetting when switching tabs
- Fixed cross-compilation targets in release pipeline (rust-toolchain.toml override)
Infrastructure
- Friendly archive names (linux-x86_64, macos-aarch64, etc.) replacing Rust target triples
- Combined checksums.sha256 file (was 5 separate files)
- Idempotent cargo publish for pipeline re-runs
- Improved proptest parser fuzz tests
- Dependency updates: pathfinding, rust-cache, codeql-action, upload/download-artifact
Install: cargo install sbom-tools or brew install sbom-tool/tap/sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.14...v0.1.16
Highlights
Source Tab v2 brings 12 new interactive features for exploring SBOM source data. Graph-aware diffing gets major correctness and feature improvements. The entire release pipeline is now fully automated with Trusted Publishing, SLSA L3 provenance, and signed tags.
What's New
TUI
- Source Tab v2 — 12 new features:
- Mouse click/scroll support in both diff and view modes
- XML syntax highlighting for raw SBOM source
- Line numbers toggle (
I) and word wrap toggle (W) - Status bar with position and percentage
- Diff change navigation (
n/N) — jump between added/removed/modified nodes - Copy JSON path to clipboard (
c) - Export source content to file (
E) - Regex search mode (
Ctrl+Rduring search) - Bookmarks (
mto toggle,'to cycle) - Tree filter by value type (
f) and sort by key (S) - Detail panel in diff mode (
d)
- Components and Dependencies tabs: visual polish, clipboard support, auto-expand roots
Quality & Compliance
- Software complexity index in quality metrics
Security
- Signed git tags (SSH ed25519)
- SLSA Build Level 3 provenance on GitHub Releases (
.intoto.jsonl) - Trusted Publishing (OIDC) for crates.io — no long-lived API tokens
- OSV scanner config to suppress false positives from SBOM test fixtures
- Branch protection hardened (required review thread resolution)
Bug Fixes
- Fixed UTF-8 panics in Source tab when viewing binary-like content
- Fixed vulnerability score inflating quality rating when no vuln data exists
- Fixed CI concurrency: only cancel in-progress on PRs, not main pushes
Infrastructure
- GitHub Releases with auto-generated notes and SLSA provenance attached
- PR-based release script compatible with branch protection
- Community docs: CODE_OF_CONDUCT, CONTRIBUTING, GOVERNANCE, ARCHITECTURE
- OpenSSF Scorecard improvements (Signed-Releases check fixed)
- Dependency updates: anyhow, quick-xml, chrono, spdx, ctrlc
Install: cargo install sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.10...v0.1.14
What's Changed
- Fix release.sh: let CI create GitHub Release by @matrosov in https://github.com/sbom-tool/sbom-tools/pull/40
- Release v0.1.13 by @matrosov in https://github.com/sbom-tool/sbom-tools/pull/41
Full Changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.12...v0.1.13
Highlights
- Source Tab v2 — 12 new interactive features for SBOM source exploration
- Graph-Aware Diffing — Major correctness and feature improvements
- Supply Chain Security — Signed releases, SLSA L3 provenance, OpenSSF hardening
- Community Docs — CODE_OF_CONDUCT, CONTRIBUTING, GOVERNANCE, ARCHITECTURE
What's New
Interactive TUI
- Mouse click/scroll support in Source tab (both diff and view modes)
- XML syntax highlighting for raw SBOM source
- Line numbers toggle (
I) and word wrap toggle (W) - Status bar with position/percentage
- Diff change navigation (
n/N) — jump between added/removed/modified nodes - Copy JSON path to clipboard (
c) - Export source content to file (
E) - Regex search mode (
Ctrl+Rduring search) - Bookmarks (
mto toggle,'to cycle) - Tree filter by value type (
f) and sort by key (S) - Detail panel in diff mode (
d)
Graph Diffing
- Fixed component ID normalization for cross-format matching
- Proper EdgeKey with sorted endpoints for undirected graph support
- Content-based graph hashing (xxh3) for structural comparison
- Reparenting detection and cyclic dependency depth limits
- Graph section in JSON, Markdown, HTML, CSV, and summary reports
- Multi-SBOM graph support in
diff-multicommand --graph-diff-onlyand--graph-depthCLI options
Quality & Compliance
- Software complexity index in quality metrics
Security & Infrastructure
- Signed git tags (SSH ed25519)
- SLSA Build Level 3 provenance on GitHub Releases
- GitHub Releases with auto-generated notes
- OSV scanner config to suppress SBOM fixture false positives
- Branch protection tightened (required review thread resolution)
- OpenSSF Scorecard improvements (Signed-Releases check fixed)
Community
- CODE_OF_CONDUCT.md (Contributor Covenant v2.1)
- CONTRIBUTING.md (development guide with DCO)
- GOVERNANCE.md (BDFL model, access continuity)
- ARCHITECTURE.md (module structure, data flow, design decisions)
- SECURITY.md updated with coordinated disclosure policy
Bug Fixes
- Fixed UTF-8 panics in Source tab when viewing binary-like content
- Fixed rustfmt formatting in source tab files
Install
cargo install sbom-tools
Full Changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.10...v0.1.11
Highlights
Redesigned quality scoring engine (v2.0) with smarter weight handling and hard caps. New continuous monitoring mode watches SBOM directories for changes. Several CLI/TUI usability improvements.
What's New
Commands & CLI
watchmode with xxh3 hashing, debounce, and graceful shutdown- Fail-on-vuln exit codes for CI integration
- Dry-run mode for safe testing
- Session persistence across TUI restarts
TUI
- Quality tab updated for scoring v2.0 (adaptive layout, white value text)
- General UX polish and doctest fixes
Quality & Compliance
- Scoring engine v2.0: N/A-aware weight renormalization for lifecycle metrics
- Hard caps: EOL components cap at D (Security/CRA), dependency cycles cap at B, missing hashes cap at C (Security)
- New metrics: HashQuality, DFS cycle detection, union-find component islands, BFS dependency depth, Provenance, Auditability
Security
- time 0.3.47 security fix (RUSTSEC dependency update)
Infrastructure
- MSRV bumped to 1.88
- Hardened crates.io publish workflow with OIDC Trusted Publishing
- 164 clippy lint fixes
Install: cargo install sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/compare/v0.1.9...v0.1.10
Highlights
First public release of sbom-tools — a semantic SBOM diff and analysis tool. Supports CycloneDX 1.4–1.6 and SPDX 2.2–2.3 with 9 output formats, interactive TUI, and 9 compliance standards.
What's New
Commands & CLI
diff— semantic SBOM comparison with graph-aware dependency diffingview— interactive single-SBOM explorervalidate— compliance validation against NTIA, CRA, FDA, NIST SSDF, EO 14028quality— 8-category quality scoring (Completeness, Identifiers, Licenses, Vulnerabilities, Dependencies, Integrity, Provenance, Lifecycle)query— search components across multiple SBOMs- 9 output formats: JSON, SARIF, Markdown, HTML, CSV, Summary, Table, Side-by-side, TUI
- Multi-standard validation:
--standard ntia,fda,cra,ssdf,eo14028 - Enrichment flags with EOL breakdown in overview panel
TUI
- Components tab with stats panel, type grouping, and search
- Vulnerabilities tab with severity badges, remediation info, detail panel, and 7000+ item support
- Dependencies tab with name resolution, smart truncation, and detail panels
- Quality tab with 8-bar chart and integrity/provenance panels
- Compliance tab with severity filtering
- Source tab with tree/raw views and synced dual-panel diff
- Tab-aware exports with shared export dialog and sticky status
Quality & Compliance
- 8-category quality scoring with 6 profiles (Minimal, Standard, Security, LicenseCompliance, CRA, Comprehensive)
- 9 compliance standards with SARIF output (30+ rules)
- CRA gap checks: freshness, license tracking, vuln statement, supply chain, integrity
- EOL status charts in tree stats panels
Security
- End-of-life detection via endoflife.date API
- OpenVEX integration for vulnerability exploitability context
- OSV/KEV vulnerability enrichment
Infrastructure
- Edition 2024, MSRV 1.88
- CI: lint, MSRV, 4-platform test matrix, cargo-deny, security audit
- 38 integration tests, 700+ unit tests
Install: cargo install sbom-tools
Crate: https://crates.io/crates/sbom-tools
Full changelog: https://github.com/sbom-tool/sbom-tools/commits/v0.1.9