What does this skill do, and when should you use it?
This Agent Skill is one entry in the NVIDIA/skills monorepo and focuses specifically on bootstrapping and troubleshooting the catheter_navigation workflow. It checks Linux host requirements, NVIDIA GPU and CUDA compatibility, and PYTHONPATH, then verifies CLI registration and runs CPU smoke tests. Users with a suitable GPU can also generate a synthetic DRR image for a quick sanity check. The Skill declares Apache-2.0, while the catalog repository contains 324 skills and is maintained through daily synchronization from product repositories.
It resolves or clones the i4h-workflows repository, locates workflows/catheter_navigation, adds its simulation directory to PYTHONPATH, and creates a timestamped run directory with logs. It runs command -v, nvidia-smi, and df -h preflight checks, calls ./i4h modes catheter_navigation to verify CLI registration, runs seven CPU unit tests, and can invoke render_drr to produce a DRR image. It then checks smoke.log for OK and verifies that fluorosim imports successfully.
- A developer setting up catheter_navigation for the first time can verify the repository path, CLI registration, and Python import configuration.
- A user seeing a fluorosim import failure can restore PYTHONPATH and confirm that the simulation directory exists.
- A user troubleshooting slangpy, CUDA, or GPU failures can check host requirements and choose local execution or Docker.
- A maintainer without a GPU can run the CPU smoke tests for CI-style verification.
How do you install this skill?
- Before execution, confirm permission to clone an unpinned GitHub repository and review the target paths, network access, and writes under $HOME.
- This skill verifies setup but does not fully install dependencies; GPU or dependency failures still require supported-version and Docker guidance.
- The default flow depends on GitHub reachability, which is not shown to be reliable from mainland-China networks.
- Do not interpret passing CPU smoke tests as evidence that GPU render_drr or interactive_viewport is ready.
- Shell / CLI
- Network access
- Local filesystem
Linux x86_64Ubuntu 22.04 or 24.04 testedGitPython 3numpyNVIDIA GPU with compute capability >= 7.0CUDA 12.8-compatible driverslangpyWarptorch CUDADocker optional
In an Agent Skills-compatible client, run: npx skills add nvidia/skills --skill i4h-catheter-navigation-setup --yes. The README does not document a separate installer for this Skill. Using it also requires an i4h-workflows checkout containing workflows/catheter_navigation, or permission for the documented commands to clone the repository to ~/i4h-workflows.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/i4h-catheter-navigation-setup" ~/.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?
Once installed, send your agent any of these to trigger it:
- Set up catheter_navigation
- Troubleshoot catheter_navigation fluorosim, GPU, or slangpy errors.
Prompt the Agent with “Set up catheter_navigation” or “Troubleshoot catheter_navigation fluorosim, GPU, or slangpy errors.” The Skill resolves the workflow from I4H_WORKFLOWS or the current Git root; if it is not found, it uses ~/i4h-workflows and clones https://github.com/isaac-for-healthcare/i4h-workflows. It then performs the CLI registration check and CPU smoke tests in order; with a GPU and slangpy installed, it can additionally run render_drr.
What are this skill's strengths and limitations?
- States concrete host, GPU, CUDA, memory, and disk requirements.
- Provides copyable commands for CLI registration, CPU smoke testing, and optional DRR validation.
- CPU tests do not require a GPU, supporting CI-style verification.
- Includes troubleshooting paths for PYTHONPATH, repository-root, and slangpy/GPU failures.
- There is no dedicated setup.sh; the Skill primarily verifies and documents requirements.
- Full GPU modes require an NVIDIA GPU, CUDA, slangpy, Warp, and torch CUDA.
- The documented tested environment is limited to Linux x86_64 with Ubuntu 22.04/24.04.
- No independent performance benchmark or broader platform-test 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 Setup Verification this page ✓ NVIDIA · Official | 51 · 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 Fluoroscopy Smoke Tests ✓ NVIDIA · Official | 50 · 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 |
| i4h Catheter Navigation Workflow Guide ✓ NVIDIA · Official | 52 · Use with care | ★ 3.8k | today | Apache-2.0 |
How did FollowSkills review this skill?
The document states host scope, credential expectations, and avoids obvious destructive commands; however, it may clone an external repository without confirmation, write under $HOME, create symlinks, and run environment commands. The dependency is not commit-pinned, rollback is unspecified, and network/file data flow is incompletely disclosed, so points are deducted.
The ordered steps, path variables, expected seven-test result, and troubleshooting guidance are useful; however, behavior depends on the current contents of an external repository, dependencies and Docker versions are not pinned, failure feedback is partial, and execution cannot be confirmed statically. The score therefore stays below 10.
Activation scenarios, Linux/NVIDIA requirements, CPU versus GPU boundaries, and non-fit routing away from digital-twin work are reasonably clear; however, the environment is narrow, Chinese-language support is unspecified, and the default GitHub clone has no demonstrated mainland-China reachability, so points are deducted.
The skill provides frontmatter, version, Apache-2.0 licensing, ownership, prerequisites, limitations, troubleshooting, and final-output guidance; however, the benchmark identifies a potentially ignored root version field, recommended Instructions and Examples sections are absent, and changelog plus explicit maintenance/update paths are missing.
It can directly guide host checks, CLI registration validation, CPU smoke tests, optional GPU checks, and reusable logs; however, it does not perform installation, GPU dependency remediation is high-level, and the committed evaluation reports limited effectiveness, so points are deducted.
Concrete commands, expected outputs, log paths, and committed evaluation material provide some auditability; however, there is no independently verifiable CI, execution artifact, or reproducible external evidence, and the benchmark is largely a declarative report, so only limited credit is justified.
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 →