Test Your AMC Stack in Minutes with This Sample Calibration Skill

Where this fits

You've just deployed an NVIDIA Auto Magic Calib (AMC) microservice and need to confirm the stack actually works before pointing it at production footage. Instead of hunting for a test video, configuring paths, and guessing whether the calibration pipeline is healthy, you can run the bundled sample dataset (sdg_08_2_sample_data_010926.zip) end-to-end. This skill automates that validation.

Why agents benefit

• The agent can trigger a full calibration run on the shipped sample dataset with a single natural-language prompt, eliminating manual script setup.
• It disambiguates between sample data, custom video files, and RTSP streams, routing to the correct calibration skill (amc-run-video-calibration or amc-run-rtsp-calibration) instead of failing silently.
• It can chain with amc-setup-calibration-stack to spin up the microservice first, then immediately validate it — a complete install-to-verify loop.
• The agent returns structured results (calibration metrics, status codes, logs) that can feed into downstream checks or dashboards.

Practical scenarios

Verify a fresh install

After running npx skills add NVIDIA/skills --skill amc-run-sample-calibration, you ask "verify AMC install." The agent starts the stack, runs the sample calibration, and reports pass/fail — no manual steps.

Sanity-check before a demo

You're about to demo AMC to stakeholders. The agent runs the sample dataset to confirm the pipeline is green, so you don't discover a broken config mid-demo.

CI-style regression check

You've updated the AMC service or configuration. The agent re-runs the sample calibration as a quick regression test, catching issues before you commit to production workloads.

Add it to your agent workflow

Install the skill:

npx skills add NVIDIA/skills --skill amc-run-sample-calibration

Then prompt your agent:

"run sample calibration against the AMC microservice"

The agent will locate the sample dataset, invoke the REST API, and summarize the calibration outcome.

Compare before adopting

Before adopting, compare against:

• Quality signals: Does the skill return actionable pass/fail metrics, or just a vague "success"?
• Maintenance freshness: Check the reviewed date (2026-04-28) and whether the owner (NVIDIA) updates it alongside AMC releases.
• Alternatives: If you routinely work with custom videos or RTSP streams, the sibling skills (amc-run-video-calibration, amc-run-rtsp-calibration) may be a better fit.
• Workflow fit: Does the skill's trigger phrasing match how your team actually asks for validation?

Why it is worth tracking

This skill is a strong quality signal for NVIDIA's agent-skill ecosystem: it's Apache-2.0 licensed, actively reviewed, and backed by a 3.1k-star repository. If you're building agents that manage NVIDIA infrastructure, tracking this skill (and its siblings) keeps your validation loop aligned with upstream best practices. Evaluate it whenever you add a new AMC deployment or change calibration parameters. For a deeper look at the skill's metadata and alternatives, see the skill page.