원클릭으로 Manus에서 모든 스킬 실행

mr-synthesis-quality-v1

스타6

포크2

업데이트2026년 5월 31일 08:17

Used to verify nv-generate-mr and nv-generate-mr-brain evidence packs by re-reading generated MR NIfTI images and checking geometry, finite nonconstant nonnegative intensities, runtime identity, and scope disclosure.

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

NVIDIA-Medtech

NVIDIA-Medtech/medical-AI-skills

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

MR Synthesis Quality Verifier

Purpose

Used for deterministic second-pass review of nv-generate-mr and nv-generate-mr-brain evidence packs.
Re-reads generated image artifacts and checks source-pack success, upstream entrypoint identity, model inventory, modality/version identity, shape and spacing consistency, finite nonconstant nonnegative voxel values, aggregate-flag honesty, runtime identity, and scope disclosure.
Manifest I/O: inputs are nv_generate_mr_evidence_pack; outputs are mr_synthesis_quality_report.

Instructions

Use this verifier only on evidence packs produced by skills/nv-generate-mr or skills/nv-generate-mr-brain.
Run it through eval_engine/run.py when producing verifier evidence.
The verifier entrypoint is scripts/grade.py; do not replace it with prompt-only review.
Treat a pass as an engineering floor only, not as validation of anatomical realism, diagnostic utility, or training-data suitability.

Available Scripts

Script	Purpose	Arguments
`scripts/grade.py`	Primary verifier entrypoint declared by `skill_manifest.yaml`.	`EVIDENCE_PACK_DIR`

Prerequisites

Runtime requirements: Python packages listed in runtime.side_effects.pip_packages.
The input directory must contain manifest.json, validation_summary.json, and output.json from a supported MR generation evidence pack.
Referenced generated image artifacts must be available for NIfTI header/data checks.

Limitations

Does not validate anatomical realism, radiologic image quality, diagnosis, regulatory claims, or production training-data suitability.
Basic numeric checks catch empty/corrupt outputs, not subtle model failures.
Does not re-run the upstream diffusion model.

Troubleshooting

Error	Cause	Fix
`target_skill_matches` fails	The fixture is not an `nv-generate-mr` or `nv-generate-mr-brain` evidence pack.	Re-run the verifier against a supported source pack.
`image_artifacts_readable` fails	A generated NIfTI path is missing, moved, or unreadable.	Regenerate or restore the source evidence pack artifacts.
`image_values_nonconstant` fails	The generated image is constant or nearly empty.	Inspect upstream logs and regenerate evidence after fixing runtime/input issues.
`aggregate_flags_match_recomputed` fails	Source output flags disagree with verifier recomputation.	Fix the wrapper summary logic and regenerate the source pack.

Example

python eval_engine/run.py verifiers/mr_synthesis_quality_v1 \
  --fixture runs/nv_generate_mr_demo \
  --out runs/mr_synthesis_quality

python verifiers/mr_synthesis_quality_v1/scripts/grade.py \
  runs/nv_generate_mr_demo

이 저장소의 다른 Skills

같은 저장소

skill-completeness-v1

NVIDIA-Medtech/medical-AI-skills

Grades a skill or verifier directory for required files, valid manifest structure, side-effect declarations, validation gates, paired-verifier resolution, fixtures, and authoring hygiene. Engineering verification only.

2026-07-086

ct-segmentation-finetune-quality-v1

NVIDIA-Medtech/medical-AI-skills

Deterministic verifier for nv_segment_ct_finetune evidence packs. Audits the finetuned checkpoint, training trajectory, and dataset audit signals already recorded by the wrapper. Engineering verification only.

2026-05-316

ct-segmentation-quality-v1

NVIDIA-Medtech/medical-AI-skills

Audits nv_segment_ct evidence packs for label-map readability, anatomy plausibility (organ volume bounds, fragmentation, bilateral symmetry, liver-larger-than-spleen), and optional per-class Dice/IoU against a referenced ground-truth label map. Engineering verification only.

2026-05-316

ct-synthesis-quality-v1

NVIDIA-Medtech/medical-AI-skills

Deterministic paired verifier for nv_generate_ct_rflow evidence packs. Audits per-sample image/mask geometry, CT-HU plausibility, mask label-set sanity, declared output-label coverage, and lung-lobe HU plausibility. CPU-only; does not re-run synthesis.

2026-05-316

dicom-metadata-quality-v1

NVIDIA-Medtech/medical-AI-skills

Used to verify dicom-metadata-extract evidence packs for source gate success, header fact completeness, PHI-flag consistency, and scope disclosure. Not for de-identification.

2026-05-316

dicom-preflight-quality-v1

NVIDIA-Medtech/medical-AI-skills

Audits dicom_series_preflight evidence packs for pass/warn/fail preflight verdicts with explicit corruption, orientation, PHI, and consistency findings.

2026-05-316

name	mr-synthesis-quality-v1
description	Used to verify nv-generate-mr and nv-generate-mr-brain evidence packs by re-reading generated MR NIfTI images and checking geometry, finite nonconstant nonnegative intensities, runtime identity, and scope disclosure.
license	Apache-2.0
allowed-tools	Bash
metadata	{"author":"NVIDIA MedTech <noreply@nvidia.com>","tags":["MedTech","MR","generation","verifier"]}

MR Synthesis Quality Verifier

Purpose

Used for deterministic second-pass review of nv-generate-mr and nv-generate-mr-brain evidence packs.
Re-reads generated image artifacts and checks source-pack success, upstream entrypoint identity, model inventory, modality/version identity, shape and spacing consistency, finite nonconstant nonnegative voxel values, aggregate-flag honesty, runtime identity, and scope disclosure.
Manifest I/O: inputs are nv_generate_mr_evidence_pack; outputs are mr_synthesis_quality_report.

Instructions

Use this verifier only on evidence packs produced by skills/nv-generate-mr or skills/nv-generate-mr-brain.
Run it through eval_engine/run.py when producing verifier evidence.
The verifier entrypoint is scripts/grade.py; do not replace it with prompt-only review.
Treat a pass as an engineering floor only, not as validation of anatomical realism, diagnostic utility, or training-data suitability.

Available Scripts

Script	Purpose	Arguments
`scripts/grade.py`	Primary verifier entrypoint declared by `skill_manifest.yaml`.	`EVIDENCE_PACK_DIR`

Prerequisites

Runtime requirements: Python packages listed in runtime.side_effects.pip_packages.
The input directory must contain manifest.json, validation_summary.json, and output.json from a supported MR generation evidence pack.
Referenced generated image artifacts must be available for NIfTI header/data checks.

Limitations

Does not validate anatomical realism, radiologic image quality, diagnosis, regulatory claims, or production training-data suitability.
Basic numeric checks catch empty/corrupt outputs, not subtle model failures.
Does not re-run the upstream diffusion model.

Troubleshooting

Error	Cause	Fix
`target_skill_matches` fails	The fixture is not an `nv-generate-mr` or `nv-generate-mr-brain` evidence pack.	Re-run the verifier against a supported source pack.
`image_artifacts_readable` fails	A generated NIfTI path is missing, moved, or unreadable.	Regenerate or restore the source evidence pack artifacts.
`image_values_nonconstant` fails	The generated image is constant or nearly empty.	Inspect upstream logs and regenerate evidence after fixing runtime/input issues.
`aggregate_flags_match_recomputed` fails	Source output flags disagree with verifier recomputation.	Fix the wrapper summary logic and regenerate the source pack.

Example

python eval_engine/run.py verifiers/mr_synthesis_quality_v1 \
  --fixture runs/nv_generate_mr_demo \
  --out runs/mr_synthesis_quality

python verifiers/mr_synthesis_quality_v1/scripts/grade.py \
  runs/nv_generate_mr_demo