fix: report Keras H5 external refs despite metadata

[mldangelo-oai]

Summary

• treat model-supplied fixed Keras metadata as metadata-only when HDF5 external references are present
• emit a warning-level CVE-2026-1669 metadata check instead of a passed version check
• add regression coverage for fixed metadata with external refs and a benign fixed-metadata no-external-refs guard

Validation

• PYTHONPATH=/private/tmp/modelaudit-c151 PROMPTFOO_DISABLE_TELEMETRY=1 /Users/mdangelo/code/modelaudit/.venv/bin/pytest tests/scanners/test_keras_h5_scanner.py -q
• /Users/mdangelo/code/modelaudit/.venv/bin/ruff format --check modelaudit/scanners/keras_h5_scanner.py tests/scanners/test_keras_h5_scanner.py
• /Users/mdangelo/code/modelaudit/.venv/bin/ruff check modelaudit/scanners/keras_h5_scanner.py tests/scanners/test_keras_h5_scanner.py
• PYTHONPATH=/private/tmp/modelaudit-c151 /Users/mdangelo/code/modelaudit/.venv/bin/mypy modelaudit/scanners/keras_h5_scanner.py tests/scanners/test_keras_h5_scanner.py
• git diff --check

[mldangelo-oai]

@codex review

[github-actions]

Workflow run and artifacts

Performance Benchmarks

Compared 12 shared benchmarks with a regression threshold of 15%.
Status: 0 regressions, 0 improved, 12 stable, 0 new, 0 missing.
Aggregate shared-benchmark median: 772.12ms -> 777.25ms (+0.7%).

Workload	Benchmark	Target	Size	Files	Baseline	Current	Change	Status
mixed-model-repository	tests/benchmarks/test_scan_benchmarks.py::test_scan_release_candidate_repository	release-candidate	547.3 KiB	32	294.58ms	305.92ms	+3.8%	stable
duplicate-heavy-registry	tests/benchmarks/test_scan_benchmarks.py::test_scan_duplicate_registry_snapshot	registry-snapshot	915.2 KiB	13	221.24ms	216.78ms	-2.0%	stable
clean-training-checkpoint	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_clean_training_checkpoint	safe_large	278.2 KiB	1	19.80ms	19.42ms	-2.0%	stable
padded-multi-stream-upload	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_padded_multi_stream_upload	multi_stream_padded	4.1 KiB	1	1.54ms	1.51ms	-1.8%	stable
nested-payload-review	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_nested_payload_review[nested_raw]	nested_raw	78 B	1	406.9us	412.7us	+1.4%	stable
single-checkpoint-preflight	tests/benchmarks/test_scan_benchmarks.py::test_scan_single_checkpoint_before_load	single_checkpoint.pkl	183.0 KiB	1	40.83ms	40.36ms	-1.1%	stable
warm-cache-rescan	tests/benchmarks/test_scan_benchmarks.py::test_scan_warm_cached_repository_rescan	release-candidate	547.3 KiB	32	66.09ms	65.46ms	-0.9%	stable
chunked-upload-stream	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_chunked_upload_stream	chunked_stream	278.2 KiB	1	22.69ms	22.52ms	-0.7%	stable
nested-payload-review	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_nested_payload_review[nested_base64]	nested_base64	98 B	1	410.0us	411.2us	+0.3%	stable
nested-payload-review	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_nested_payload_review[nested_hex]	nested_hex	130 B	1	419.0us	418.0us	-0.2%	stable
suspicious-pickle-intake	tests/benchmarks/test_scan_benchmarks.py::test_scan_suspicious_pickle_intake	suspicious-intake	183.8 KiB	4	102.67ms	102.59ms	-0.1%	stable
direct-malicious-upload	tests/benchmarks/test_picklescan_benchmarks.py::test_picklescan_direct_malicious_upload	malicious_reduce	52 B	1	1.45ms	1.45ms	-0.1%	stable

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Another round soon, please!

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[ianw-oai]

Looks good.