원클릭으로 Manus에서 모든 스킬 실행

nv-reason-cxr-quality-v1

스타6

포크2

업데이트2026년 5월 31일 08:17

Used to verify nv-reason-cxr evidence packs for source-pack success, fixture/image hash binding, NV-Reason-CXR runtime identity, response shape, scope disclosure, and conservative forbidden overreach phrases.

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

NV Reason CXR Quality Verifier

Purpose

Used for deterministic second-pass review of an nv-reason-cxr evidence pack.
Checks source-pack success, fixture case/prompt binding, image readability and hash, model/runtime identity, output response shape, limitation disclosure, and forbidden overreach phrases.
Manifest I/O: inputs are nv_reason_cxr_evidence_pack; outputs are nv_reason_cxr_quality_report.

Instructions

Use this verifier only on an evidence pack directory produced by skills/nv-reason-cxr.
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 engineering evidence only, not as validation of clinical correctness or model quality.

Available Scripts

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

Prerequisites

Runtime requirements: Python standard library only.
The input directory must contain manifest.json, validation_summary.json, and output.json from an nv-reason-cxr evidence pack.
The recorded fixture path and image artifact must be available for full fixture/image binding.

Limitations

Does not validate chest X-ray interpretation, model correctness, report completeness, or medical actionability.
Forbidden-phrase checks are regex-based and can miss paraphrases.
Mock-mode packs prove wrapper and gate wiring; they do not prove live NV-Reason-CXR-3B model behavior.

Troubleshooting

Error	Cause	Fix
`target_skill_matches` fails	The fixture is not an `nv-reason-cxr` evidence pack.	Re-run the verifier against the source skill pack directory.
`fixture_loaded` fails	The source pack references a fixture that is unavailable in this checkout.	Regenerate the source pack with a committed synthetic fixture.
`image_sha256_matches` fails	The recorded image artifact changed or is unavailable.	Regenerate the source pack or inspect artifact provenance.
`forbidden_phrases_absent` fails	Output contains treatment, regulatory, patient-directed, or absolute-certainty language.	Fix the source skill prompt/runtime guard and regenerate evidence.

Example

MOCK_NV_REASON_CXR=1 python eval_engine/run.py skills/nv-reason-cxr \
  --fixture skills/nv-reason-cxr/fixtures/synthetic_cxr_input.json \
  --out runs/nv_reason_cxr_demo

python eval_engine/run.py verifiers/nv_reason_cxr_quality_v1 \
  --fixture runs/nv_reason_cxr_demo \
  --out runs/nv_reason_cxr_quality

python verifiers/nv_reason_cxr_quality_v1/scripts/grade.py \
  runs/nv_reason_cxr_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	nv-reason-cxr-quality-v1
description	Used to verify nv-reason-cxr evidence packs for source-pack success, fixture/image hash binding, NV-Reason-CXR runtime identity, response shape, scope disclosure, and conservative forbidden overreach phrases.
license	Apache-2.0
allowed-tools	Bash
metadata	{"author":"NVIDIA MedTech <noreply@nvidia.com>","tags":["MedTech","radiology","chest-xray","verifier"]}

NV Reason CXR Quality Verifier

Purpose

Used for deterministic second-pass review of an nv-reason-cxr evidence pack.
Checks source-pack success, fixture case/prompt binding, image readability and hash, model/runtime identity, output response shape, limitation disclosure, and forbidden overreach phrases.
Manifest I/O: inputs are nv_reason_cxr_evidence_pack; outputs are nv_reason_cxr_quality_report.

Instructions

Use this verifier only on an evidence pack directory produced by skills/nv-reason-cxr.
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 engineering evidence only, not as validation of clinical correctness or model quality.

Available Scripts

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

Prerequisites

Runtime requirements: Python standard library only.
The input directory must contain manifest.json, validation_summary.json, and output.json from an nv-reason-cxr evidence pack.
The recorded fixture path and image artifact must be available for full fixture/image binding.

Limitations

Does not validate chest X-ray interpretation, model correctness, report completeness, or medical actionability.
Forbidden-phrase checks are regex-based and can miss paraphrases.
Mock-mode packs prove wrapper and gate wiring; they do not prove live NV-Reason-CXR-3B model behavior.

Troubleshooting

Error	Cause	Fix
`target_skill_matches` fails	The fixture is not an `nv-reason-cxr` evidence pack.	Re-run the verifier against the source skill pack directory.
`fixture_loaded` fails	The source pack references a fixture that is unavailable in this checkout.	Regenerate the source pack with a committed synthetic fixture.
`image_sha256_matches` fails	The recorded image artifact changed or is unavailable.	Regenerate the source pack or inspect artifact provenance.
`forbidden_phrases_absent` fails	Output contains treatment, regulatory, patient-directed, or absolute-certainty language.	Fix the source skill prompt/runtime guard and regenerate evidence.

Example

MOCK_NV_REASON_CXR=1 python eval_engine/run.py skills/nv-reason-cxr \
  --fixture skills/nv-reason-cxr/fixtures/synthetic_cxr_input.json \
  --out runs/nv_reason_cxr_demo

python eval_engine/run.py verifiers/nv_reason_cxr_quality_v1 \
  --fixture runs/nv_reason_cxr_demo \
  --out runs/nv_reason_cxr_quality

python verifiers/nv_reason_cxr_quality_v1/scripts/grade.py \
  runs/nv_reason_cxr_demo