I4h Workflow E2e

NVIDIAOfficialCreated by Isaac for Healthcare TeamDistributed by NVIDIA

Skill
Workflow
Automation and integrations
Apache-2.0
Medium risk

From the author

Run the maintained workflow data-to-policy pipeline from recording through checkpoint validation. Use for full end-to-end requests; do not use for one individual stage.

— the description by Isaac for Healthcare Team

AutoGineer summary

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

The i4h-workflow-e2e skill provides a standardized driver for running end-to-end robotics workflows. It manages the entire lifecycle of a data-to-policy pipeline, ensuring consistency across stage resolution, artifact generation, and checkpoint handoff. By utilizing the maintained driver, the agent ensures that all task manifests, environment configurations, and dependencies are correctly handled from initial recording to final validation.

What it does

Resolves and prepares the required workflow repository and environment.
Performs a mandatory dry-run to verify stage paths and backend ownership.
Executes the full data-to-policy pipeline in the foreground.
Manages stage sequencing, logging, and run directory creation.
Validates artifacts including HDF5 files, LeRobot metadata, and training checkpoints.
Provides a structured summary of stage outcomes and final verification status.

When to use it

When you need to execute a full end-to-end robotics workflow.
When you require a consistent, automated pipeline from recording to checkpoint validation.
When you need to verify the integrity of a specific policy workflow.
When troubleshooting a pipeline failure by inspecting stage-specific logs.

Advantages

Ensures consistency across complex multi-stage robotics workflows.
Automates the verification of artifacts and logs for every pipeline stage.
Prevents manual configuration errors by using maintained workflow manifests.
Provides clear visibility into pipeline progress and failure points.

Usage examples

“Run end-to-end smoke pipeline for scissor pick-and-place.”: The agent performs a dry-run, executes the driver for the scissor pick-and-place workflow, and reports the status of each stage from recording to validation.

Install

Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/nvidia/i4h-workflow-e2e/download

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

mkdir -p .agents/skills && unzip i4h-workflow-e2e.zip -d .agents/skills

Documentation by Isaac for Healthcare Team

Run the Workflow End-to-End Pipeline

Purpose

Use the maintained driver so stage resolution, artifacts, logs, and checkpoint handoff stay consistent with current workflow/task manifests.

Instructions

Resolve the base checkout and policy workflow.
Require a successful driver dry-run.
Execute the maintained driver in the foreground.
Inspect every stage artifact before reporting completion.

Resolve the checkout

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"

Treat this resolver 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.

Require the workflow's policy mode. The driver discovers the remote task, embodiment, task text, and trainability from live workflow/task manifests.

Dry-run first

./scripts/e2e/run.sh --env <workflow> --dry-run

Require exit status 0 and inspect every printed command and artifact path. The dry-run is the source of truth for current stages and backend ownership.

Run in the foreground

./scripts/e2e/run.sh --env <workflow>

Use --run-dir only when the caller needs a specific location. Apply --skip-mimic, --skip-annotate, --skip-replay, or --skip-viz only when the user explicitly omits that optional stage or a documented smoke profile requires it.

Keep the driver as this agent's foreground tool call. Do not use a subagent, monitor task, shell backgrounding, nohup, tmux, or a detached process. Poll until exit.

The driver performs full setup, then owns its stage sequence, timestamped run directory, runs/.latest link, and per-stage logs. Do not replace it with a manually assembled subset.

Verify

On success, inspect the printed summary and artifacts:

policy recording
expanded/filtered HDF5 as applicable
visible replay result when enabled
LeRobot metadata, parquet, and videos
visualizer URL/content when enabled
training logs and exact checkpoint when trainable
checkpoint validation recording and final success summary

On failure, stop at the first failed stage, inspect that stage's log, preserve the run directory, and repair the owning stage before rerunning. Do not skip a required failure merely to obtain a green summary. Stop leftovers with ./stop.sh all.

Troubleshooting

Use the first failed stage and its log to choose the owning stage skill. Preserve the run directory and rerun only after that stage verifies its output.

Prerequisites

Require a policy workflow plus host, simulator, backend, VLM, dataset, training, and visualization dependencies for every enabled stage.

Limitations

The pipeline supports only workflows with a policy mode; inference-only Tasks skip fine-tuning and checkpoint validation.

Examples

Run end-to-end smoke pipeline for scissor pick-and-place. → dry-run, execute the driver, and report each recording-to-validation stage.

Completion gate

Report workflow/task/embodiment/trainability, dry-run result, run directory, every stage outcome and skip, dataset/visualizer/checkpoint/verification artifacts, final exit status, and cleanup state.

Files

ATTRIBUTION.mdAdded by AutoGineer925 B
BENCHMARK.md6.5 KiB
LICENSELicence file added by AutoGineer11.2 KiB
SKILL.md4.4 KiB
evals/evals.json1.3 KiB
skill-card.md4.4 KiB
skill.oms.sig4.5 KiB

BENCHMARK.md

Skill Benchmark: i4h-workflow-e2e

✅ Overall verdict: PASS — Recommended for publication

Publication Recommendation

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

Evaluation Metadata

Skill: i4h-workflow-e2e
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:df688f31b369fdc7f19cf4212dec493696ea1df04dd7347724dcfc0dea93a4dc (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	62.8% — baseline ran, but no comparable score was available; uplift unavailable	57.3% — baseline ran, but no comparable score was available; uplift unavailable
Security	83.3% → 33.3% (-50.0 points)	75.0% → 0.0% (-75.0 points)
Correctness	6.7% → 100.0% (+93.3 points)	25.0% → 100.0% (+75.0 points)
Discoverability	85.0% — baseline ran, but no comparable score was available; uplift unavailable	75.0% — baseline ran, but no comparable score was available; uplift unavailable
Effectiveness	6.3% → 20.2% (+13.9 points)	22.1% → 52.9% (+30.8 points)
Efficiency	75.3% — baseline ran, but no comparable score was available; uplift unavailable	58.6% — 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	4,596,116	1,249,779	N/A	N/A	skill 3/3; base 6/6
claude-code	i4h-workflow-e2e-readme-quick-scissor	4,111,889	630,613	N/A	N/A	skill 2/2; base 3/3
claude-code	i4h-workflow-e2e-table-scissor	484,227	619,166	N/A	N/A	skill 1/1; base 3/3
codex	All cases	8,028,492	2,830,886	N/A	N/A	skill 2/2; base 4/4
codex	i4h-workflow-e2e-readme-quick-scissor	3,464,765	2,524,382	+940,383	+37.25%	skill 1/1; base 1/1
codex	i4h-workflow-e2e-table-scissor	4,563,727	306,504	N/A	N/A	skill 1/1; base 3/3
ALL AGENTS	Dataset aggregate	12,624,608	4,080,665	N/A	N/A	skill 5/5; base 10/10

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); 1 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 SECURITY/Unknown (SQP-2): The skill silently clones a remote repository into the user's home directory (e.g., $HOME/i4h-workflows) when I4H_WOR (SKILL.md:31`)

</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-workflow-e2e · 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-workflow-e2e.zip