Holoscan Install Source

NVIDIAOfficialCreated by Holoscan TeamDistributed by NVIDIA

Skill
Workflow
Software development
Apache-2.0
Medium risk

From the author

Build and verify Holoscan SDK 5.x from an explicit source ref with the repository run wrapper; use for Early Access or SDK development.

— the description by Holoscan Team

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

This skill provides a structured procedure for compiling and validating the Holoscan SDK 5.x from specific source references. It is designed for developers working with Early Access features, custom components, or debug builds. The workflow covers environment verification, source repository management, build configuration using the SDK's internal run wrapper, and post-build verification through targeted testing.

What it does

Verifies system prerequisites including CUDA, Docker, and build tools.
Manages source code checkouts for specific tags, commits, or branches.
Executes build processes using the repository's native run wrapper.
Supports custom build configurations, including debug modes and architecture targeting.
Validates the build through automated C++ and Python test execution.
Provides guidance on integrating the built SDK into consumer projects.

When to use it

When you need to build unpublished Early Access features of the Holoscan SDK.
When developing custom components or modifications for the SDK.
When a debug build is required for troubleshooting.
When working with specific SDK versions or branches not available in pre-built packages.

Advantages

Ensures build consistency by utilizing the repository's integrated run wrapper.
Provides clear steps for verifying system dependencies before starting the build.
Supports flexible build configurations for different architectures and requirements.
Includes verification steps to confirm successful compilation and functionality.

Usage examples

“Build the Holoscan SDK 5.x from the main branch.”: The agent will clone the repository, verify system dependencies, and execute the build command using the default configuration.
“Perform a debug build of the Holoscan SDK for ARM64.”: The agent will configure the build process with the debug flag and target the ARM64 architecture, followed by the compilation step.

Install

Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/nvidia/holoscan-install-source/download

Unzip it into the cross-client skills folder of your project:

mkdir -p .agents/skills && unzip holoscan-install-source.zip -d .agents/skills

Documentation by Holoscan Team

Build Holoscan 5.x from Source

Choose the source and prerequisites

Use source for unpublished Early Access features, SDK changes, custom components, or debug builds. This workflow uses the checkout's build container with Ubuntu 24.04 and CUDA 13, targeting x86_64 or arm64 dGPU. Check Linux, disk capacity, GPU/driver, Git, Docker, buildx, and NVIDIA Container Toolkit availability. Explain the container download and build cost before starting.

Select the explicit 5.x source ref from the release notes or the user's requested development branch. Inspect an existing checkout before reusing it; keep its branch and local changes intact. Resolve the requested ref to its full commit, using the intended remote's ref for a remote branch and fetching that ref if necessary. From the existing repository root, compare its HEAD with the resolved ref before building:

git status --short
git rev-parse HEAD
git rev-parse --verify "${SDK_REF}^{commit}"

SDK_REF must identify the selected tag, commit, or remote-qualified branch. If the ref cannot be resolved or the commits differ, stop and report the mismatch. Preserve the existing branch and working tree; offer a separate checkout at the requested ref. Report any local changes even when HEAD matches, and preserve them when the user requested building their modified checkout.

For a new public checkout:

# SDK_REF is a confirmed published 5.x tag or development branch.
git ls-remote --heads --tags https://github.com/nvidia-holoscan/holoscan-sdk.git
mkdir -p ~/holoscan
git clone --branch "$SDK_REF" https://github.com/nvidia-holoscan/holoscan-sdk.git ~/holoscan/holoscan-sdk-5x
cd ~/holoscan/holoscan-sdk-5x
cat VERSION
./run --help
./run build --help

Verify VERSION is 5.x. Read this checkout's AGENTS.md, DEVELOP.md, and docs/sdk_installation.mdx; these govern an unreleased branch. In an internal checkout the exported source root is the directory containing VERSION and run; run the commands from that directory. Never assume the repository's default branch or a 4.x example is the requested 5.x source.

Build

./run check_system_deps
./run build --dryrun
./run build

The wrapper builds, configures, compiles, and installs. It enables Python, examples, operator testing, and optional modules by default. Useful choices:

Option	Meaning
`--type debug` / `--type rel-debug`	Debug / release with debug information
`--arch arm64`	ARM64 target; consult the checkout's cross-compilation prerequisites
`--buildpath build-custom`	Explicit build directory
`--installprefix install-custom`	Explicit install prefix
`--build-python false`	C++-only build
`--config "-DHOLOSCAN_BUILD_VISION=OFF"`	Disable an optional component when supported by the checkout

Use the same target selection for build, directory queries, and launch. The standard trees are build-<arch> and install-<arch>; use ./run get_build_dir and ./run get_install_dir rather than the 4.x build-cu<N>-<arch> naming. Custom paths must be passed explicitly. Start with --reconfigure for stale configuration; inspect what a cleanup command removes before using it.

Verify

For the default build directory:

./run test --name 'EXAMPLE_CPP_HELLO_WORLD_TEST|EXAMPLE_CPP_HELLO_CORE_TEST|EXAMPLE_PYTHON_FIRST_PIPELINE_TEST'

Confirm the selected tests were found and executed, rather than treating zero tests as success. A C++-only build omits the Python case. For a custom directory:

./run launch build-custom --run-cmd "ctest -N"
./run launch build-custom --run-cmd "ctest -R 'EXAMPLE_CPP_HELLO_WORLD_TEST|EXAMPLE_CPP_HELLO_CORE_TEST|EXAMPLE_PYTHON_FIRST_PIPELINE_TEST' --output-on-failure"

These cover greeting, bounded C++ Core execution, and the 5.x Python pipeline. Run additional module/GPU tests appropriate to the selected components; state what remains untested. Cross-compilation does not qualify execution on the target architecture.

Use the absolute install prefix as CMAKE_PREFIX_PATH for consumer projects using find_package(holoscan) and holoscan::core. Report the source commit, build options, build/install directories, and test results. For GPU passthrough or buildx failures, report the failing command and diagnostic output, then resolve the prerequisite identified by the checkout's checks. If the cause remains unknown, stop before building and request the missing host information instead of retrying indefinitely or changing unrelated system settings.

Files

ATTRIBUTION.mdAdded by AutoGineer934 B
BENCHMARK.md8.2 KiB
LICENSELicence file added by AutoGineer11.2 KiB
SKILL.md4.8 KiB
evals/evals.json7.0 KiB
skill-card.md4.7 KiB
skill.oms.sig4.5 KiB

BENCHMARK.md

Skill Benchmark: holoscan-install-source

✅ Overall verdict: PASS — Recommended for publication

Publication Recommendation

Recommended for publication based on the completed evaluation evidence in this report.

Evaluation Metadata

Skill: holoscan-install-source
Evaluation date: 2026-10-15
Evaluator version: 1.5.6
Agents: Claude Code (aws/anthropic/bedrock-claude-opus-4-8), Codex (openai/openai/gpt-5.5)
Tasks: 6 evaluation tasks (5 positive, 1 negative)
Dataset digest: sha256:8a153da53159948d77200bc314d7c3d7e7dba6b6df680fcd0eb31f5109fdb051 (skill-evaluator-dataset-snapshot/1)
Attempts per task: 1
Environment: k8s-sandbox
Tier 2 evidence: required for publication
Tier 3 evidence: required for publication

Each task attempt ran in its own isolated sandbox pod.

What This Report Answers

The three-tier evaluation checks whether the skill:

is safe to use;
produces correct answers;
is discovered and activated when needed;
helps the agent complete the user's goal and expected workflow; and
avoids wasted skill and tool usage.

Results at a Glance

Measure	Claude Code (Baseline → Skill Uplift)	Codex (Baseline → Skill Uplift)
Overall	97.7% — baseline ran, but no comparable score was available; uplift unavailable	97.1% — baseline ran, but no comparable score was available; uplift unavailable
Security	66.7% → 100.0% (+33.3 points)	83.3% → 100.0% (+16.7 points)
Correctness	83.3% → 100.0% (+16.7 points)	100.0% → 96.7% (-3.3 points)
Discoverability	100.0% — baseline ran, but no comparable score was available; uplift unavailable	95.0% — baseline ran, but no comparable score was available; uplift unavailable
Effectiveness	65.8% → 90.0% (+24.2 points)	78.3% → 95.0% (+16.7 points)
Efficiency	98.5% — baseline ran, but no comparable score was available; uplift unavailable	98.7% — baseline ran, but no comparable score was available; uplift unavailable

How to read this table: baseline is the same task attempted without the target skill. Scores are rounded to one decimal; threshold-adjacent values use additional precision so their displayed band matches the verdict. Uplift is derived from those displayed scores and shown in percentage points.

Example: 47.0% → 92.0% (+45.0 points) means the skill-assisted run scored 92.0%, 45.0 percentage points above its 47.0% no-skill baseline.

A partial dimension was calculated from only the available configured signals; review the detailed report before relying on it.

Token Usage

Actual Tier 3 execution usage is reported for every observed agent/case pair and both conditions.

Agent	Dataset case	With skill	Without skill	Delta	Change	Coverage
claude-code	All cases	477,734	1,539,621	-1,061,887	-68.97%	skill 6/6; base 6/6
claude-code	holoscan-install-source-5x-001	62,817	426,384	-363,567	-85.27%	skill 1/1; base 1/1
claude-code	holoscan-install-source-5x-002	62,766	184,777	-122,011	-66.03%	skill 1/1; base 1/1
claude-code	holoscan-install-source-5x-003	62,846	186,712	-123,866	-66.34%	skill 1/1; base 1/1
claude-code	holoscan-install-source-5x-004	162,349	431,392	-269,043	-62.37%	skill 1/1; base 1/1
claude-code	holoscan-install-source-5x-005	63,402	154,090	-90,688	-58.85%	skill 1/1; base 1/1
claude-code	holoscan-install-source-5x-006	63,554	156,266	-92,712	-59.33%	skill 1/1; base 1/1
codex	All cases	238,954	383,825	-144,871	-37.74%	skill 6/6; base 6/6
codex	holoscan-install-source-5x-001	29,589	13,979	+15,610	+111.67%	skill 1/1; base 1/1
codex	holoscan-install-source-5x-002	28,973	13,816	+15,157	+109.71%	skill 1/1; base 1/1
codex	holoscan-install-source-5x-003	29,432	14,251	+15,181	+106.53%	skill 1/1; base 1/1
codex	holoscan-install-source-5x-004	92,332	313,651	-221,319	-70.56%	skill 1/1; base 1/1
codex	holoscan-install-source-5x-005	29,033	14,145	+14,888	+105.25%	skill 1/1; base 1/1
codex	holoscan-install-source-5x-006	29,595	13,983	+15,612	+111.65%	skill 1/1; base 1/1
ALL AGENTS	Dataset aggregate	716,688	1,923,446	-1,206,758	-62.74%	skill 12/12; base 12/12

Prompt tokens include cached reads, so total tokens are prompt + completion (cached is not added twice). The Efficiency score uses (prompt - cached) + completion. N/A means the relevant trajectory counters were not available; coverage is never estimated.

Tier Status

Tier	Purpose	Status	Evidence
Tier 1	Static validation	PASSED WITH OBSERVATIONS	11 validator(s); 7 finding(s)
Tier 2	Semantic deduplication	PASSED	2 validator(s); 0 finding(s)
Tier 3	Live agent evaluation	PASS	2 agent(s); 6 task(s)

Findings and Observations

<details> <summary>Show detailed findings and successful checks</summary>

MEDIUM SCHEMA/body_recommended_section: Missing recommended section: '## Instructions' (team-skills/holoscan/holoscan-sdk/holoscan-install-source/SKILL.md)
MEDIUM SCHEMA/body_recommended_section: Missing recommended section: '## Examples' (team-skills/holoscan/holoscan-sdk/holoscan-install-source/SKILL.md)
LOW QUALITY/quality_discoverability: No '## Purpose' section (team-skills/holoscan/holoscan-sdk/holoscan-install-source/SKILL.md)
LOW QUALITY/quality_reliability: No prerequisites/requirements documented (team-skills/holoscan/holoscan-sdk/holoscan-install-source/SKILL.md)
LOW QUALITY/quality_reliability: No limitations documented (team-skills/holoscan/holoscan-sdk/holoscan-install-source/SKILL.md)
2 additional finding(s) are available in the full evaluation artifacts.

</details>

Scoring Methodology

<details> <summary>Show dimension definitions, source signals, and thresholds</summary>

Dimension	Question	Scored signals
Security	Is it safe to use?	`security` (100%)
Correctness	Is the answer correct?	`accuracy` (100%)
Discoverability	Was the right skill loaded when needed?	`skill_execution` (100%)
Effectiveness	Did the skill help complete the task?	`goal_accuracy` (50%) + `behavior_check` (50%)
Efficiency	Did it avoid wasted tool calls and token usage?	`skill_efficiency` (50%) + `token_efficiency` (50%)

Dimension bands: PASS at 50% or above; NEUTRAL from 40% to below 50%; FAIL below 40%.
Overall Tier 3 lift: PASS at +5 points or more; FAIL at -10 points or less; values between those bands are NEUTRAL.
Overall verdict: PASS only when every configured dimension passes for at least one supported agent. Lift is reported as diagnostic evidence and does not override this gate.
The 50% attempt pass threshold is a separate per-task gate; it is not the dimension pass threshold.
Effectiveness is the equal-weight mean of goal completion (goal_accuracy) and expected workflow adherence (behavior_check).
Efficiency is 50% tool-call productivity (the backward-compatible skill_efficiency wire id) and 50% token_efficiency. Positive-case skill routing is scored under Discoverability, not Efficiency; a negative case without a routing target is N/A. N/A sources are omitted, remaining weights are renormalized, and the dimension is marked partial.

Signals present in this run:

security (Security): unsafe operations, secret leakage, and unauthorized access.
skill_execution (Skill Execution): whether the expected skill was selected, decoys were avoided, and the workflow executed.
skill_efficiency (Tool Productivity): tool-call productivity (legacy wire id; routing is scored under Discoverability).
accuracy (Accuracy): final-answer correctness against the reference answer.
goal_accuracy (Goal Accuracy): whether the user's goal was achieved.
behavior_check (Behavior Check): whether the expected workflow behavior was followed.
token_efficiency (Token Efficiency): actual uncached prompt plus completion usage (50% of Efficiency).

</details>

Freshness

Regenerate this benchmark when the skill, evaluation dataset, target agent/model, evaluator version, environment, or scoring policy changes.

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Needs: Needs a command-line tool, Needs Python

Security

What our scanner noticed· Medium risk

Large base64-encoded blobs

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Created by

Holoscan Team

Publisher

NVIDIA (Official)

Distributed by

NVIDIA

Source

NVIDIA/skills · skills/holoscan-install-source · commit fce3cca

Licence

Apache License 2.0 (redistribution allowed)

LICENSE (added by AutoGineer from LICENSE-APACHE)
ATTRIBUTION.md (written by AutoGineer)

Synced

2026-11-08

AutoGineer is not affiliated with or endorsed by Holoscan Team or NVIDIA.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download holoscan-install-source.zip