What does this skill do, and when should you use it?
This skill is part of NVIDIA/skills and targets the i4h catheter-navigation workflow; its version is 0.7.0. It renders one DRR fluoroscopy frame from a preprocessed CT cache, a direct NIfTI/DICOM path, or a built-in synthetic phantom. Actual rendering requires an NVIDIA GPU and slangpy. It is limited to single-frame rendering rather than batched multi-environment reinforcement-learning rendering.
Resolves or prepares the i4h-workflows directory, creates a timestamped run directory with logs, and invokes ./i4h run catheter_navigation render_drr. Without --cache it uses the synthetic phantom; with a cache it checks for mu_volume.npy, renders drr.png, and records output in render_drr.log. The skill states that direct NIfTI/DICOM paths are supported, but does not document a corresponding command example.
- An engineer smoke-testing the Slang renderer or GPU setup can render the synthetic phantom without input data.
- A medical-imaging developer with a preprocessed CT cache can generate a single DRR frame from a directory containing mu_volume.npy.
- A catheter-navigation team can use the frame as a static smoke test for its rendering workflow.
- A user troubleshooting a run can inspect the generated PNG and log file as verification artifacts.
How do you install this skill?
- Before execution, confirm that the remote clone, output directories, and symlink target are authorized.
- Do not expose patient CT cache data, paths, or logs in prompts, shared outputs, or remote services; add retention and cleanup guidance.
- GitHub dependency may be unreachable from mainland China or restricted environments; provide a preseeded repository, mirror, or offline installation path.
- Move the version into the schema-recognized metadata field and add Instructions, Examples, pinned dependencies, and reproducible tests.
- Shell / CLI
- Network access
- Local filesystem
NVIDIA GPUslangpyCUDA 12.8NVIDIA driver >=570Git
Install the individual skill with the skills CLI:
npx skills add nvidia/skills --skill i4h-catheter-navigation-render-drr --yes
The README also supports selecting a target agent, such as by adding --agent codex. After installation, the skill is available when the agent next loads skills and encounters a relevant task. If i4h-workflows is not present locally, the documented setup can clone it from https://github.com/isaac-for-healthcare/i4h-workflows.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/i4h-catheter-navigation-render-drr" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Complete i4h-catheter-navigation-setup first. For the synthetic phantom:
./i4h run catheter_navigation render_drr --local --run-args="--output ${OUTPUT}"
For a CT cache, set CACHE to a preprocess_ct output directory containing mu_volume.npy, then run:
./i4h run catheter_navigation render_drr --local --run-args="--cache ${CACHE} --output ${OUTPUT}"
Verify the result with test -f "${OUTPUT}" and file "${OUTPUT}". The final report should include the PNG path, whether synthetic or cache input was used, and the log path.
What are this skill's strengths and limitations?
- Provides a no-data synthetic-phantom smoke test.
- Renders a single DRR from a preprocessed CT cache.
- Defines run-directory, output-file, and log verification steps.
- Documents both local and Docker-oriented troubleshooting paths.
- Actual rendering depends on an NVIDIA GPU, slangpy, CUDA, and a compatible driver environment.
- It only renders one frame and does not cover batched multi-environment reinforcement-learning rendering.
- The default example does not attach a CatheterProvider, so catheter compositing is not included by default.
- No concrete command is documented for direct NIfTI/DICOM input, and no test-suite or platform-coverage evidence is provided.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Catheter Navigation DRR Renderer this page ✓ NVIDIA · Official | 52 · Use with care | ★ 3.8k | today | Apache-2.0 |
| i4h Catheter Navigation Workflow Guide ✓ NVIDIA · Official | 52 · Use with care | ★ 3.8k | today | Apache-2.0 |
| Catheter Navigation Fluoroscopy Viewport ✓ NVIDIA · Official | 50 · Use with care | ★ 3.8k | today | Apache-2.0 |
| Catheter Navigation E2E Smoke Test ✓ NVIDIA · Official | 45 · Not recommended | ★ 3.8k | today | Apache-2.0 |
| Catheter Navigation Setup Verification ✓ NVIDIA · Official | 51 · Use with care | ★ 3.8k | today | Apache-2.0 |
Compared with CT-cache rendering, synthetic-phantom mode requires no data and is suited to fast smoke tests; cache mode is suited to rendering from preprocessed CT data. The skill also identifies Docker mode without --local as an alternative troubleshooting path for local slangpy/CUDA failures.
How did FollowSkills review this skill?
The skill distinguishes synthetic data from CT cache use and does not require credentials; however, it executes git clone, creates directories and a symlink, may process patient cache data, and lacks confirmation, least-privilege guidance, data-flow disclosure, sensitive-data safeguards, and rollback, so points are deducted.
The ordered workflow, cache-file check, output verification, and GPU/CUDA troubleshooting are reasonably clear; however, it depends on an external repository, GPU, slangpy, and unfixed versions, with no key-path tests in the selected skill, so static evidence is limited.
Trigger phrases, synthetic/cache scenarios, single-frame scope, and separation from the interactive viewport are clear; however, Chinese-language support is not addressed, core setup depends on GitHub and other potentially unreachable overseas services, and environment-fit evidence is limited, so points are deducted.
The skill provides a name, version, description, license, owner, dependencies, limitations, troubleshooting, evaluation report, and references; however, the root-level version may be ignored by the schema, recommended Instructions and Examples sections are missing, and maintenance ownership/update procedures are underspecified.
The intended PNG and log outputs are explicit, with both data-free synthetic and CT-cache paths; the benchmark reports partial success on positive tasks. Static review cannot independently verify those results, and GPU dependence plus single-frame scope limit direct utility, so points are deducted.
The material includes a pinned revision context, evaluation tasks, behavioral criteria, and summarized results, providing limited auditability; however, it lacks committed tests or reproducible CI covering this skill’s key paths, and the benchmark claims cannot be independently reproduced from the supplied files.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →