Generated banner for Omniverse Usd Performance Tuning
Generated logo for Omniverse Usd Performance Tuning

Omniverse Usd Performance Tuning

NVIDIAOfficialCreated by NVIDIA Omniverse

  • Skill
  • Workflow
  • Software development
  • Apache-2.0
  • Medium risk
  • Has scripts

From the author

Diagnose and optimize USD stages for slow loading, high memory, low FPS or validation; preserve asset boundaries and reuse shared definitions.
— the description by NVIDIA Omniverse

AutoGineer summary

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

This skill provides a structured workflow for diagnosing and optimizing USD stages. It focuses on improving performance metrics like load times, memory footprint, and frame rates by identifying and optimizing unique asset definitions, reusing shared definitions, and pruning unnecessary data. The process emphasizes preserving original asset boundaries, animation, and interface requirements, ensuring that downstream applications remain functional after optimization.

What it does

  • Inspects USD stages to identify performance bottlenecks and structural inefficiencies.
  • Classifies assets into atomic, package, or aggregate structures to guide optimization.
  • Performs native USD optimization operations on disjoint scopes.
  • Publishes semantic libraries to enable efficient asset reuse.
  • Validates composition, ownership, and interface accessibility of the optimized output.
  • Profiles the final output against the original baseline to report measured improvements.

When to use it

  • When a USD stage suffers from slow loading times or high memory consumption.
  • When rendering performance (FPS) needs improvement in a production environment.
  • Before finalizing a USD assembly for downstream consumption.
  • When you need to standardize asset structures according to factory-level guidelines.

Advantages

  • Maintains asset boundaries and critical metadata during optimization.
  • Reduces scene complexity through intelligent reuse of shared definitions.
  • Provides verifiable reports on performance gains and composition integrity.
  • Supports large-scale scene optimization by processing independent definition scopes.

Usage examples

“Inspect the USD stage at /data/factory.usdc for performance issues and prepare a review workspace.”
The agent initializes the workspace, runs the inspection script, and provides a summary of candidates for optimization.
“Submit the optimization decisions for the factory stage based on the generated JSON.”
The agent processes the boundary decisions, executes the optimization operations, and generates the final optimized USD assembly.

Install

  1. Download the skill as a ZIP:

    https://autogineer.com/agent-skills/nvidia/omniverse-usd-performance-tuning/download
  2. Unzip it into the cross-client skills folder of your project:

    mkdir -p .agents/skills && unzip omniverse-usd-performance-tuning.zip -d .agents/skills

Documentation by NVIDIA Omniverse

Omniverse USD Performance Tuning

Purpose

Improve loading, memory use or rendering for the requested workload. Preserve the prim paths, appearance, placement and behavior required by downstream applications. Follow the component/subcomponent model in NVIDIA's Factory-Level USD Structuring guide. Apply the relevant parts of that model; a static asset does not need new animation or physics.

Inputs

Required: the source USD stage. When no workload is specified, use factory layout planning and visualization and record that assumption in the request/report. Explicit workload and preservation requirements take precedence. Optional: an output workspace, preservation requirements, physical error budget, existing library indexes and recorded runtime evidence. Use values from the current user request; reuse recorded values only while they remain applicable. Follow setup when no output path is supplied.

Prerequisites

Verify the selected runtime and required capabilities using setup. Keep its preflight record with the run; installation and remote access are conditional steps described there.

Instructions

  1. Inspect. Establish the workload, preservation requirements and output workspace. Set up progress and dashboard access before assessment. Verify setup and inspect the selected source configuration.

  2. Judge boundaries. Review equipment, service-system sections and architectural elements in the selected configuration, including animation and interiors. Record owners, editable interfaces, equipment categories and optional families using boundary guidance.

  3. Classify and author. Classify accepted assets children before parents. Capture or reuse the original profile before authoring. Prepare reusable definitions, materials and occurrence opinions using asset authoring.

  4. Optimize unique definitions. Run supported native operations concurrently on disjoint scopes with the recorded physical budget.

  5. Publish semantic libraries. Use one writer per category layer and remap consumer references.

  6. Check composition. Check authored ownership, references, bindings, placement, animation and interface accessibility through the written-output checks.

  7. Profile and report. Compare the final consumer with the original baseline under comparable conditions and report measured benefits, completion and operations by asset definition.

sequenceDiagram
    participant A as Agent
    participant T as USD tools
    participant W as Definition workers
    participant L as Library publisher
    participant D as Read-only progress dashboard
    Note over A,L: One compact progress state also renders text progress
    T-->>D: Recorded phase, assessment rounds, counters and decision reasons
    A->>T: Inspect original selected configuration
    T->>T: Collect cached child-to-parent summaries and repetition nominations
    T->>T: Load supplied indexes once and batch authored product clues
    T-->>A: Product review groups, member context, animation and reuse evidence
    loop Until boundaries and interiors are reviewed
        A->>A: Select asset boundaries, then identify required internal prims and properties
        A->>T: Submit boundary decisions with reviewed scope, required edits, category and optional family
        T->>T: Reuse unchanged source evidence, recompute boundary-dependent results
        opt Geometry ownership requires reconstruction
            T->>T: Regroup or native split, remap paths and reassess affected regions
        end
        T-->>A: Remaining candidates and member exceptions
    end
    T->>T: Classify selected child assets before parent assets, confirm reuse
    T->>T: Capture or reuse original baseline profile
    T->>T: Confirm accepted product identities from source and supplied indexes; reconcile complete parent-child groups
    T->>T: Skip worker inputs for confirmed groups; author remaining materials, editable prims and occurrence overrides
    T->>T: Determine merge groups on prepared interiors; preserve selected child assets and required prims
    T->>W: Read-only input USD layers for existing-library matching
    par Independent preparation jobs
        W->>W: Confirm remaining product, exact source or authorized normalized reuse; preserve occurrence placement
        W->>W: Consolidate accepted product or compatible exact/normalized leaf inputs; retain every occurrence
    end
    W-->>T: Confirmed library hits and definitions requiring optimization
    T->>W: Owned geometry prim paths in read-only input layers
    par Independent definition scopes
        W->>W: Remove small meshes when authorized; apply asset representation: reserve merge inputs or fit analytic prims; reduce, merge (native clustering if required), prune, flatten, prune
    end
    W->>L: Geometry outputs and assembly authoring outcomes
    L->>L: Share eligible static whole definitions within current category publication
    L->>L: Share remaining whole meshes; write USD reuse metadata and compact index
    L->>T: Output USD stage with rewritten references and dependencies
    T->>T: Check authored composition, ownership, animation and interfaces
    T->>T: Profile final output USD stage and report results by asset definition
    T-->>A: Measured changes, coverage and unresolved work
    T-->>D: Final verdict and report/library paths

Use the provided tools for traversal, native operations, authoring and checks; the agent supplies engineering judgments. Preserve current variant selections, authored animation, required materials and interfaces. Trust supported USD Optimize operations with valid settings and scopes. Deep geometric comparisons and whole-scene edit trials belong in development tests, not ordinary execution.

Complete when every unique definition has an executed or evidenced unchanged outcome, every asset has justified ownership and structure, and the written consumer passes composition checks. Retain native failures, unresolved judgments and missing required capabilities as incomplete work. Repeat only affected work when evidence changes.

Scope and authorization

  • Diagnosis and validation requests remain read-only. An optimization request authorizes work in a separate output workspace within its stated preservation limits.

  • Never overwrite source layers or inherited libraries without explicit authorization.

  • Reuse authorization already given. Ask only for a missing decision that changes identity, required appearance or a loss budget; explain the affected component and consequence.

  • Report only observed changes. Measure rendering before claiming higher FPS or lower GPU use.

Load only the relevant guidance

RequestRead
Validation onlySetup, then the validation section of operations
Approved operation chainSetup, operations, verification
Existing or persistent asset libraryAdd the library index requirements to the assessment
Missing standalone runtimeInstallation
Remote stageRemote access before the first open
Report from completed evidenceVerification

Use upstream package guidance for selected operation mechanics and proxy workflows. Keep large logs on disk as specified in setup.

Limitations

Convert CAD input with the source converter's workflow and NVIDIA's CAD conversion guidance. Render profiling requires NVIDIA/omniperf or the user's available profiler; USD open and traversal timings alone do not establish rendering improvements.

Troubleshooting

For unavailable imports, operations or remote assets, follow setup. For incomplete execution or failed composition checks, follow verification; retain the failed artifacts and report the missing capability or failed check.

Examples

After runtime preflight, use the verified Python interpreter from setup to inspect a stage:

python scripts/successor_assessment.py inspect --source /data/factory.usdc --workspace /data/factory-review

Review the generated candidates using boundary guidance, fill the generated decision template, then submit it:

python scripts/successor_assessment.py decide --workspace /data/factory-review --decisions /data/factory-review/decisions.json

These commands perform assessment; continue through the instructions above only when review is complete. Run from the skill directory; generated data stays in the workspace.

Output Format

Maintain live progress and editable steering throughout execution.

For end-to-end optimization, provide the written USD assembly, validated report JSON, and generated Markdown and HTML reports. State the runtime, operations actually executed, preservation and coverage results, measured changes, source-write status, and artifact paths. Diagnosis-only work may return a findings summary stating that no optimized stage was written.

Files

scripts/successor_progress.py

#!/usr/bin/env python3
# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
"""Render inexpensive progress and accepted decision summaries; never inspect USD.

Usage:
    Use Progress(manifest, directory) as a context manager in successor_workflow.
    Send phase events to event() and the final outcome to finish(). No CLI.
Arguments:
    The assessed manifest, output directory and workflow event dictionaries.
Output:
    Writes progress summaries and steering.md in the run directory; preserves
    existing user edits to steering.md. Reads worker receipts without opening USD.
Exit codes:
    Not applicable. Progress snapshot I/O failures are logged and retried on the next update.
"""
import collections
import json
import importlib.util
import logging
from pathlib import Path
import threading
import time



# One presentation map for Markdown, snapshots and the browser. Unlisted tool
# substeps remain within the current workflow phase.
STAGES = [('assessment', 'Assess asset structure'), ('source_validation', 'Validate source'),
          ('source_profile', 'Profile source'), ('author_inputs', 'Prepare assets'),
          ('persistent_matching', 'Confirm library reuse'), ('definition_compute', 'Optimize assets'),
          ('library_publication', 'Publish libraries'), ('composed_verification', 'Verify assembly'),
          ('output_profile', 'Profile output'), ('report', 'Report results')]
PHASE_STAGE = {key: key for key, _ in STAGES}
PHASE_STAGE.update(starting='assessment', product_confirmation='author_inputs',
    product_confirmation_complete='author_inputs', definition_compute_complete='definition_compute',
    library_publication_complete='library_publication', composed_verification_complete='composed_verification',
    complete='report')
ACTIVITIES = {
    'assessment': 'Starting assessment', 'stage_open': 'Opening the selected source stage',
    'scan_stage': 'Collecting source hierarchy and boundary evidence',
    'source_evidence_reused': 'Reusing collected source evidence',
    'boundary_review': 'Preparing candidate boundary groups',
    'boundary_review_complete': 'Candidate groups ready for agent review',
    'inventory': 'Inspecting geometry and material assignments',
    'exact_confirmation': 'Confirming candidate reuse',
    'assessment_complete': 'Asset assessment complete',
    'complete': 'Execution finished',
}
PHASES = {key: (title, 'Continue the recorded workflow') for key, title in STAGES}
PHASES.update({key: (title, 'Continue the recorded workflow') for key, title in ACTIVITIES.items()})


def read_state(path):
    """Read only a compact snapshot; never load a manifest or USD for progress."""
    try:
        path = Path(path)
        if path.stat().st_size > 8 * 1024 * 1024:
            return {}
        value = json.loads(path.read_text(encoding='utf-8'))
        return value if isinstance(value, dict) else {}
    except (OSError, ValueError):
        return {}


def counter(phase, facts, workers):
    if phase == 'source_evidence_reused':
        return None
    for done, total, unit in [('prims_done', 'prims_total', 'prims visited in this activity'),
                             ('completed_count', 'definition_count', 'assets processed'),
                             ('completed', 'total', 'items processed')]:
        if isinstance(facts.get(done), (int, float)):
            return {'done': facts[done], 'total': facts.get(total), 'unit': unit}
    overall = workers.get('overall', {})
    if phase in ('persistent_matching', 'definition_compute', 'definition_compute_complete') and overall:
        return {'done': sum(overall.get(k, 0) for k in ('done', 'failed', 'timeout', 'blocked')),
                'total': overall.get('total'), 'unit': 'jobs finished'}
    return None


def decision_rows(manifest):
    groups = collections.Counter()
    examples = {}
    for definition in manifest.get('definitions', []):
        selection, rationale = definition.get('selection', {}), definition.get('rationale', {})
        key = (definition.get('semantic_category') or 'general',
               definition.get('structural_type') or 'Structure not yet assigned',
               selection.get('reason') or rationale.get('reason') or 'Reason not recorded',
               selection.get('independent_use') or '', definition.get('equipment_family') or '')
        groups[key] += 1
        examples.setdefault(key, definition.get('boundary_path', definition.get('id', '')))
    return [dict(category=k[0], structure=k[1], reason=k[2], independent_use=k[3], family=k[4],
                 count=n, example=examples[k]) for k, n in groups.most_common()]


def decision_summary(manifest, limit=12):
    groups = collections.Counter()
    examples = {}
    for definition in manifest.get('definitions', []):
        selection = definition.get('selection', {})
        rationale = definition.get('rationale', {})
        reason = selection.get('reason') or rationale.get('reason') or 'Reason not recorded; review required.'
        key = (definition.get('semantic_category', 'general'), definition.get('structural_type', 'unclassified'),
               selection.get('basis', 'recorded'), reason, selection.get('independent_use', ''), definition.get('equipment_family'))
        groups[key] += 1
        examples.setdefault(key, definition.get('boundary_path', definition.get('id', '')))
    lines = []
    for key, count in groups.most_common(limit):
        category, structure, basis, reason, use, family = key
        lines.append(f'- **{category}: {structure}** — {count} definition(s); {basis}. {reason}'
                     + (f' Family: {family}.' if family else '')
                     + (f' Independent use: {use}.' if use else '')
                     + f' Example: `{examples[key]}`.')
    if len(groups) > limit:
        lines.append(f'- {len(groups)-limit} further decision groups are recorded in the assessment manifest.')
    return '\n'.join(lines) or '- No accepted definition decisions are available yet.'


def operation_decisions(manifest, limit=8):
    choices = collections.Counter()
    for definition in manifest.get('definitions', []):
        for row in definition.get('assessment', {}).get('operations', []):
            if isinstance(row, dict):
                choices[(row.get('category', 'operation'), row.get('status', 'unrecorded'), row.get('reason', 'See the recorded operation scope.'))] += 1
    lines = [f'- {category} ({status}): {reason} ({count} definition assessment(s)).'
             for (category, status, reason), count in choices.most_common(limit)]
    if len(choices) > limit:
        lines.append(f'- {len(choices)-limit} further operation groups are recorded in the authored manifest.')
    return '\n'.join(lines)


class Progress:
    def __init__(self, manifest, directory, *, assessment=False, common_library=None):
        self.directory = Path(directory)
        self.directory.mkdir(parents=True, exist_ok=True)
        self.started = time.monotonic()
        self.started_epoch = time.time()
        self.assessing = assessment
        self.stage = 'assessment' if assessment else 'source_validation'
        self.phase_snapshots = {}
        self.events = []
        self.rounds = []
        self.rows = []
        self.common_library = str(Path(common_library).resolve()) if common_library else None
        self.phase_started = self.started
        self.assessment_pass = None
        self.source_evidence_reused = False
        self.phase = 'starting'
        self.state = 'running'
        self.facts = {}
        self.workers = {}
        self.active_operations = []
        self.stop = threading.Event()
        self.lock = threading.RLock()
        self.decisions = decision_summary(manifest)
        self.operation_choices = operation_decisions(manifest)
        self.source = manifest.get('original_source', manifest.get('source', {})).get('path', 'unspecified')
        previous = read_state(self.directory/'progress.json')
        if not previous and manifest.get('workspace'):
            previous = read_state(Path(manifest['workspace'])/'progress.json')
        if previous.get('source') == self.source:
            self.started_epoch = previous.get('started_at', self.started_epoch)
            self.phase_snapshots = previous.get('phases', {})
            self.events = previous.get('events', [])
            self.rounds = previous.get('assessment_rounds', [])
            self.rows = previous.get('decisions', [])
            self.decisions = previous.get('decisions_markdown', self.decisions)
            if assessment:
                self.phase_started = self.started - max(0, time.time()-self.started_epoch)
        if manifest.get('definitions'):
            self.rows = decision_rows(manifest)
            self.decisions = decision_summary(manifest, limit=len(manifest['definitions']))
        request = manifest.get('request', {})
        path = Path(__file__).resolve().parents[1]/'references/geometry_tolerance.py'
        spec = importlib.util.spec_from_file_location('progress_geometry_budget', path)
        budget = importlib.util.module_from_spec(spec); spec.loader.exec_module(budget)
        self.requirements = f"Workload: {request.get('workload', 'factory visualization')}. Native tolerance: {budget.physical_tolerance_mm(request)} mm."
        choices = [f'{key}={request[key]}' for key in ('preservation', 'preservation_profile', 'lossless',
                   'exact_geometry', 'fit_primitives', 'decimate', 'preserve_normals', 'preserve_uvs') if key in request]
        if choices:
            self.requirements += ' Recorded constraints: '+', '.join(choices)+'.'
        protected = set(request.get('protected_paths', []))
        protected.update(path for definition in manifest.get('definitions', []) for path in definition.get('protected_paths', []))
        self.scope = f"Accepted scope: {len(manifest.get('definitions', []))} unique definitions; {len(protected)} protected interface paths."
        if protected:
            self.scope += ' Examples: '+', '.join('`'+path+'`' for path in sorted(protected)[:3])+'.'
        self.scope += ' Keep the complete scope and constraints in [the authored manifest](authored-manifest.json) unless explicitly revised.'
        self.steering = self.directory/'steering.md'
        try:
            with self.steering.open('x', encoding='utf-8') as stream:
                stream.write('# Steer the next run\n\n'
                    f'Optimize `{self.source}`. {self.requirements}\n'+self.scope+'\n'
                    'Use the selected variant configuration. Preserve authored animation, materials, '
                    'required interfaces, source layers and inherited libraries.\n\n'
                    '## Decisions to keep or revise\n\n'+self.decisions+'\n\n'
                    '## Changes for the next run\n\n'
                    '- [Edit: name the asset group, desired change and reason.]\n\n'
                    'Review this text and submit it as the next prompt. Edits here are preserved; '
                    'the running workflow does not execute this file.\n')
        except FileExistsError:
            pass
        self.render()
        self.thread = threading.Thread(target=self._watch, daemon=True)

    def __enter__(self):
        self.thread.start()
        return self

    def __exit__(self, kind, error, traceback):
        if error is not None:
            self.finish(False, 'interrupted' if issubclass(kind, (KeyboardInterrupt, SystemExit)) else 'failed', str(error) or kind.__name__)
        self.stop.set()
        self.thread.join(timeout=2)
        self.render()

    def event(self, phase, **facts):
        with self.lock:
            if 'assessment_pass' in facts:
                self.assessment_pass = facts['assessment_pass']
            if phase == 'source_evidence_reused':
                self.source_evidence_reused = True
            if phase != self.phase:
                if phase in ('persistent_matching', 'definition_compute'):
                    self.workers = {}
                    self.active_operations = []
            if not self.assessing and phase in PHASE_STAGE and PHASE_STAGE[phase] != self.stage:
                self.snapshot()  # Close the preceding phase at this transition.
            self.phase, self.facts = phase, facts
            if not self.assessing and phase in PHASE_STAGE:
                stage = PHASE_STAGE[phase]
                if stage != self.stage:
                    self.phase_started = time.monotonic()
                self.stage = stage
            row = {'phase': phase, 'activity': ACTIVITIES.get(phase, phase.replace('_', ' ')),
                   'time': time.time(), 'facts': facts}
            if self.events and self.events[-1]['phase'] == phase:
                self.events[-1] = row
            else:
                self.events.append(row)
                self.events = self.events[-100:]
            if self.assessing and self.assessment_pass is not None:
                if not self.rounds or self.rounds[-1]['number'] != self.assessment_pass:
                    self.rounds.append({'number': self.assessment_pass, 'steps': [], 'status': 'running'})
                steps = self.rounds[-1]['steps']
                item = dict(row, counter=counter(phase, facts, {}))
                if steps and steps[-1]['phase'] == phase:
                    steps[-1] = item
                else:
                    steps.append(item)
            self.render()

    def finish(self, passed, status, error=''):
        with self.lock:
            self.state = ('assessed' if passed else 'waiting') if self.assessing and status in ('assessed', 'boundary review required') else ('passed' if passed else status or 'incomplete')
            if self.assessing and self.rounds:
                self.rounds[-1]['status'] = self.state
            self.facts = {**self.facts, 'passed': passed, 'status': status}
            if error:
                self.facts['error'] = error
            self.render()

    def _watch(self):
        stamp = None
        last = time.monotonic()
        while not self.stop.wait(1):
            try:
                folder = 'persistent-matching' if self.stage == 'persistent_matching' else 'compute'
                summary = self.directory/folder/'status.summary.json'
                current = (folder, summary.stat().st_mtime_ns) if summary.exists() else None
                changed = current != stamp
                if changed:
                    data = json.loads(summary.read_text())
                    with self.lock:
                        self.workers = data
                    stamp = current
                operations = []
                for job in self.workers.get('running_jobs', []):
                    try:
                        operation = json.loads(Path(job['progress_path']).read_text())
                        if operation.get('state') == 'running':
                            operations.append(dict(operation, definition_id=job['id']))
                    except (OSError, ValueError):
                        continue
                with self.lock:
                    changed = changed or operations != self.active_operations
                    self.active_operations = operations
                if changed or time.monotonic()-last >= 30:
                    self.render()
                    last = time.monotonic()
            except (OSError, ValueError):
                # An incomplete status write must not change the execution verdict.
                continue

    def update_decisions(self, manifest):
        with self.lock:
            self.rows = decision_rows(manifest)
            self.decisions = decision_summary(manifest, limit=max(1, len(manifest.get('definitions', []))))
            self.operation_choices = operation_decisions(manifest)

    def snapshot(self):
        now = time.time()
        workers = self.workers if self.stage in ('persistent_matching', 'definition_compute') else {}
        result = {'schema_version': 1, 'updated_at': now, 'started_at': self.started_epoch,
                  'source': self.source, 'state': self.state, 'phase': self.phase, 'stage': self.stage,
                  'activity': ACTIVITIES.get(self.phase, PHASES.get(self.phase, (self.phase.replace('_', ' '),))[0]),
                  'requirements': self.requirements, 'elapsed_seconds': now-self.started_epoch,
                  'tool_elapsed_seconds': time.monotonic()-self.started,
                  'phase_elapsed_seconds': time.monotonic()-self.phase_started,
                  'facts': self.facts, 'counter': counter(self.stage if workers else self.phase, self.facts, workers),
                  'workers': workers, 'active_operations': self.active_operations if workers else [],
                  'assessment_pass': self.assessment_pass, 'assessment_rounds': self.rounds,
                  'decisions': self.rows, 'decisions_markdown': self.decisions,
                  'events': self.events, 'common_library': self.common_library}
        # Phase summaries contain facts, never copies of history or decisions.
        self.phase_snapshots[self.stage] = {key: result[key] for key in
            ('stage', 'phase', 'state', 'activity', 'updated_at', 'elapsed_seconds',
             'phase_elapsed_seconds', 'facts', 'counter', 'workers', 'active_operations')}
        result['phases'] = self.phase_snapshots
        return result

    def render(self):
        # Dashboard readers can temporarily lock a snapshot on Windows. Keep
        # the previous complete snapshot and retry at the next normal update.
        with self.lock:
            try:
                self._render()
            except OSError as error:
                if not getattr(self, '_snapshot_write_failed', False):
                    logging.getLogger(__name__).warning('Progress snapshot update failed; workflow continues: %s', error)
                self._snapshot_write_failed = True
            else:
                self._snapshot_write_failed = False

    def _render(self):
        with self.lock:
            snapshot = self.snapshot()
            target = self.directory/'progress.json'
            temporary = target.with_suffix('.json.tmp')
            temporary.write_text(json.dumps(snapshot, ensure_ascii=False), encoding='utf-8')
            temporary.replace(target)
            now = time.monotonic()
            activity, following = snapshot['activity'], 'Continue the recorded workflow'
            if self.phase == 'author_inputs' and self.facts.get('reused_authored_inputs'):
                activity = 'Reusing authored materials and definition inputs'
            if self.state not in ('running', 'passed', 'waiting', 'assessed'):
                following = 'Review the unresolved work and edit steering.md before retrying'
            elif self.state in ('passed', 'assessed'):
                following = 'Review the recorded result before continuing'
            lines = ['# Optimization progress', '', f'**Status: {self.state}.** {activity}.',
                f"Elapsed including assessment and agent review: {snapshot['elapsed_seconds']:.1f} s; current phase: {snapshot['phase_elapsed_seconds']:.1f} s.",
                f'Next: {following}.', '', self.requirements, '', f'Source: `{self.source}`.']
            if self.assessment_pass is not None:
                lines += ['', f'Assessment pass: {self.assessment_pass}.']
            if self.source_evidence_reused:
                lines += ['Source evidence reused; boundary-dependent results are recomputed.']
            if self.phase == 'scan_stage' and 'prims_done' in self.facts:
                total = self.facts.get('prims_total')
                lines += ['', f'Stage prims visited: {self.facts["prims_done"]:,}'
                          + (f' / {total:,}.' if total is not None else '; total not yet known.')]
            if self.facts:
                lines += ['', 'Current result: '+ '; '.join(f"{key.replace('_', ' ')}: {str(value)[:400] if value is not None else 'not reported'}" for key,value in self.facts.items())+'.']
            counts = snapshot['workers'].get('overall', {})
            if counts:
                lines += ['', '## Definition '+('preparation' if self.phase == 'persistent_matching' else 'optimization')+' jobs', '',
                    '; '.join(f'{name}: {counts.get(name, 0)}' for name in ('total', 'done', 'running', 'queued', 'failed', 'timeout', 'blocked'))+'.']
            if self.active_operations and self.state == 'running' and self.stage == 'definition_compute':
                lines += ['', 'Native operations currently running:', '']
                lines += [f"- {row['operation']}: {max(0, time.time()-row['started_epoch']):.1f} s elapsed; definition `{row['definition_id']}`."
                          for row in self.active_operations]
            last = self.workers.get('last_operation')
            if last:
                lines += ['', f'Last completed native operation: {last}.']
            outcomes = self.workers.get('operation_outcomes', {})
            if outcomes:
                lines += ['', 'Observed native operation outcomes (completed jobs):', '']
                lines += [f'- {operation}: '+', '.join(f'{state} {count}' for state,count in states.items())+'.'
                          for operation,states in sorted(outcomes.items())]
            reasons = self.workers.get('unchanged_reasons', {})
            if reasons:
                lines += ['', 'Recorded reasons for unchanged work:', '']+[f'- {reason} ({count} definition outcome(s)).' for reason,count in reasons.items()]
            failures = self.workers.get('failures', [])
            if failures:
                lines += ['', 'Failures requiring attention:', '']+[f'- {row}' for row in failures[:5]]
            if self.operation_choices:
                lines += ['', '## Planned operation choices', '', self.operation_choices]
            lines += ['', '## Accepted asset decisions', '', self.decisions, '',
                'These are recorded requirements and engineering judgments; operation counts above are observed results.', '',
                '## Steer another run', '', '[Edit the next-run prompt](steering.md). Progress updates preserve your edits.', '',
                'Details: [assessment events](assessment-progress.jsonl), [execution events](workflow-progress.jsonl), [worker status](compute/status.json), '
                '[authored manifest](authored-manifest.json), [final report](report.md). '
                'Files appear as their phases complete.', '']
            target = self.directory/'progress.md'
            temporary = target.with_suffix('.md.tmp')
            temporary.write_text('\n'.join(lines), encoding='utf-8')
            temporary.replace(target)

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 Python, Needs Node.js

The author says: Requires a verified standalone Usd Optimize/OpenUSD runtime for operations, usd-validation-nvidia for supported checks, and a writable output workspace. Remote assets require a working resolver. Render profiling is an external handoff.

Security

What our scanner noticed· Medium risk

  • Large base64-encoded blobs
  • Scripts that your agent may run

This skill includes scripts that your agent may run. Read them before you use it.

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
NVIDIA Omniverse
Publisher
NVIDIA (Official)
Source
NVIDIA/skills · skills/omniverse-usd-performance-tuning · commit fce3cca
Licence
Apache License 2.0 (redistribution allowed)
  • LICENSE (added by AutoGineer from LICENSE-APACHE)
  • ATTRIBUTION.md (written by AutoGineer)
Synced

AutoGineer is not affiliated with or endorsed by NVIDIA Omniverse or NVIDIA.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download omniverse-usd-performance-tuning.zip