I4h Lerobot Viz

NVIDIAOfficialCreated by Isaac for Healthcare TeamDistributed by NVIDIA

Skill
Tool integration
Data and AI
Apache-2.0
Medium risk

From the author

Serve and visually inspect a converted LeRobot dataset in the browser. Use for videos and state/action timelines; do not use for raw workflow HDF5 or incomplete conversion output.

— the description by Isaac for Healthcare Team

AutoGineer summary

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

This skill provides a mechanism to serve and visually validate converted LeRobot datasets. It enables users to inspect episode videos, state and action timelines, and metadata to ensure the dataset is correctly formatted and complete. It is designed for post-conversion verification and requires a completed dataset with a valid meta/info.json file.

What it does

Locates and identifies converted LeRobot datasets.
Launches a local web server to host the dataset for inspection.
Provides a browser-based interface to view camera videos and data timelines.
Verifies episode counts and task information against metadata.
Manages server lifecycle, including startup and cleanup commands.

When to use it

After converting a dataset to the LeRobot format.
When you need to visually verify that episode videos and timelines are rendering correctly.
To confirm that metadata matches the expected dataset content.
Before proceeding with downstream tasks that require a verified dataset.

Advantages

Enables direct visual confirmation of dataset integrity.
Provides clear feedback on state and action timeline accuracy.
Simplifies the verification process through a browser-based interface.
Includes automated cleanup instructions to manage local server resources.

Usage examples

“Open the latest converted LeRobot dataset for inspection.”: The agent identifies the most recent dataset, launches the local visualization server, provides the access URL, and summarizes the observed videos and timelines.

Install

Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/nvidia/i4h-lerobot-viz/download

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

mkdir -p .agents/skills && unzip i4h-lerobot-viz.zip -d .agents/skills

Documentation by Isaac for Healthcare Team

Visualize a LeRobot Dataset

Purpose

Serve one completed local dataset and verify its episode videos and timelines in a browser.

Instructions

Resolve the base checkout and one completed LeRobot dataset.
Launch its managed local server.
Open the printed URL.
Inspect videos, timelines, episode count, and cleanup state.

Resolve the dataset

export I4H_WORKFLOWS_REPO_URL="${I4H_WORKFLOWS_REPO_URL:-https://github.com/isaac-for-healthcare/i4h-workflows}"
I4H_REPO_DIR_NAME="${I4H_WORKFLOWS_REPO_URL%/}"
I4H_REPO_DIR_NAME="${I4H_REPO_DIR_NAME##*/}"
I4H_REPO_DIR_NAME="${I4H_REPO_DIR_NAME##*:}"
I4H_REPO_DIR_NAME="${I4H_REPO_DIR_NAME%.git}"
[ -n "$I4H_REPO_DIR_NAME" ] || { echo "Cannot derive a checkout name from I4H_WORKFLOWS_REPO_URL" >&2; exit 2; }
ROOT="${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>/dev/null)}"
if [ ! -d "$ROOT/workflows/i4h_workflows" ]; then
  ROOT="${I4H_WORKFLOWS:-$HOME/$I4H_REPO_DIR_NAME}"
  [ -d "$ROOT/workflows/i4h_workflows" ] || git clone "$I4H_WORKFLOWS_REPO_URL" "$ROOT"
fi
export I4H_WORKFLOWS="$ROOT"
cd "$ROOT"
find runs "${HF_LEROBOT_HOME:-$HOME/.cache/huggingface/lerobot}" \
  -name info.json -path '*/meta/*' -printf '%T@ %h\n' 2>/dev/null \
  | sort -nr | head

Treat the resolver above as part of the skill contract: a hosted copy may run outside the base repository, so never assume the current checkout contains workflows/i4h_workflows. I4H_WORKFLOWS_REPO_URL selects the clone source. When I4H_WORKFLOWS is unset, derive the fallback directory from that URL; set I4H_WORKFLOWS only to reuse or choose a specific destination. Never replace an existing checkout.

Use the explicit/current-chain converted directory. Otherwise select the newest candidate and state the choice. Require <dataset>/meta/info.json; route raw HDF5 to i4h-workflow-dataset-convert.

Serve

Pass an absolute local path:

DATASET_DIR=/absolute/path/to/lerobot/dataset
STATE_DIR=/absolute/path/to/run/viz-state
tools/dataset/scripts/viz.sh "$DATASET_DIR" --state-dir "$STATE_DIR"

The script selects a free local port, waits for HTTP readiness, and prints the URL, PID, state files, log, and exact stop command. Keep it running only while the user wants access.

Verify visually

Open the printed URL. Confirm:

the requested episode list loads
every expected camera video renders
state and action timelines render with plausible dimensions and motion
episode count and task text match meta/info.json
no blank page, missing video, or wrong dataset is being served

Reuse a live server only when its target dataset matches. Otherwise stop it using its printed state directory and port, then start the requested dataset.

Troubleshooting

If startup or the page fails, inspect meta/info.json, videos, the printed server log, and port ownership before restarting.

Prerequisites

Require the dataset tool environment and a completed LeRobot dataset with meta/info.json.

Limitations

The visualizer does not accept raw workflow HDF5 or repair incomplete metadata/videos.

Examples

Open the latest converted LeRobot dataset for inspection. → choose the newest verified dataset, serve it, and report the observed videos/timelines plus cleanup command.

Completion gate

Report dataset path, repo id, local URL, episode/camera/timeline observations, PID/state directory, and exact cleanup command.

Files

ATTRIBUTION.mdAdded by AutoGineer922 B
BENCHMARK.md7.0 KiB
LICENSELicence file added by AutoGineer11.2 KiB
SKILL.md3.8 KiB
evals/evals.json1.1 KiB
skill-card.md4.4 KiB
skill.oms.sig4.5 KiB

BENCHMARK.md

Skill Benchmark: i4h-lerobot-viz

✅ Overall verdict: PASS — Recommended for publication

Publication Recommendation

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

Evaluation Metadata

Skill: i4h-lerobot-viz
Evaluation date: 2026-09-21
Evaluator version: 1.5.6
Agents: Claude Code (aws/anthropic/bedrock-claude-opus-4-8), Codex (openai/openai/gpt-5.5)
Tasks: 2 evaluation tasks (2 positive)
Dataset digest: sha256:93ad31732f601b7ae680a17624f8f898357d09fcbc3494f057ea035e226ab28c (skill-evaluator-dataset-snapshot/1)
Attempts per task: 3
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	81.7% — baseline ran, but no comparable score was available; uplift unavailable	74.2% — baseline ran, but no comparable score was available; uplift unavailable
Security	100.0% → 100.0% (±0.0 points)	100.0% → 66.7% (-33.3 points)
Correctness	35.0% → 100.0% (+65.0 points)	45.0% → 100.0% (+55.0 points)
Discoverability	95.0% — baseline ran, but no comparable score was available; uplift unavailable	93.3% — baseline ran, but no comparable score was available; uplift unavailable
Effectiveness	6.3% → 16.3% (+10.0 points)	2.5% → 23.3% (+20.8 points)
Efficiency	97.3% — baseline ran, but no comparable score was available; uplift unavailable	87.8% — 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.

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	464,741	1,073,861	N/A	N/A	skill 2/2; base 4/4
claude-code	i4h-lerobot-viz-latest-converted	270,017	858,215	N/A	N/A	skill 1/1; base 3/3
claude-code	i4h-lerobot-viz-table-scissor	194,724	215,646	-20,922	-9.70%	skill 1/1; base 1/1
codex	All cases	506,817	385,235	N/A	N/A	skill 3/3; base 4/4
codex	i4h-lerobot-viz-latest-converted	307,376	254,345	N/A	N/A	skill 2/2; base 3/3
codex	i4h-lerobot-viz-table-scissor	199,441	130,890	+68,551	+52.37%	skill 1/1; base 1/1
ALL AGENTS	Dataset aggregate	971,558	1,459,096	N/A	N/A	skill 5/5; base 8/8

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); 4 finding(s)
Tier 2	Semantic deduplication	PASSED	2 validator(s); 0 finding(s)
Tier 3	Live agent evaluation	PASS	2 agent(s); 2 task(s)

Findings and Observations

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

MEDIUM QUALITY/quality_efficiency: Instructions lack clear action verbs (team-skills/holoscan/i4h-workflows/i4h-lerobot-viz/SKILL.md)
MEDIUM SECURITY/Unknown (SDI-1): The skill is described as a local dataset visualization tool, yet its core resolver logic unconditionally clones a remot (SKILL.md:33)
MEDIUM SECURITY/Unknown (SQP-2): The skill's description, purpose statement, and prerequisites make no mention of the fact that it may clone a remote git (SKILL.md:33)
MEDIUM SECURITY/Unknown (SDI-2): The clone destination is derived entirely from the I4H_WORKFLOWS_REPO_URL environment variable and defaults to $HOME/ (SKILL.md:42`)

</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.

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

Isaac for Healthcare Team

Publisher

NVIDIA (Official)

Distributed by

NVIDIA

Source

NVIDIA/skills · skills/i4h-lerobot-viz · 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 Isaac for Healthcare Team or NVIDIA.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download i4h-lerobot-viz.zip