Update Omniverse USD performance tuning skill

[jensjebens]

Onboarding type

• New product onboarding (new components.d/<slug>.yml file)
• Other (catalog change, README fix, infrastructure, etc.)

For new product onboarding — author affirmations

By submitting this PR, I confirm on behalf of my team:

• Skills cleared for open source release per NVIDIA's internal IP review process (six-question check, all answers affirmative)
• License selected: Apache 2.0 / CC-BY 4.0 / Dual (Apache 2.0 + CC-BY 4.0). Specify: Existing Omniverse manual catalog entry
• No new license or new third-party component introduced beyond what the source repo already carries
• Source repo is public and under an NVIDIA-owned GitHub org
• .agents/skills/ or skills/ path used for new entries (or existing path retained for legacy entries per components.d/<slug>.yml)

NVIDIA contributors: see the internal onboarding guide for the IP review process details and license selection.

Reviewer checklist (OSS Skills PIC)

• Author confirmations above are checked
• components.d/<slug>.yml entry valid (required fields, unique catalog_dir, path exists in source repo, filename slug matches name)
• SKILL.md frontmatter spec-compliant (at least one sampled)
• No new license or third-party dependency requiring OSRB filing

All PRs

• All commits signed off with DCO (git commit -s).
  If you forgot, run git rebase --signoff origin/main && git push --force-with-lease to retroactively sign all commits in your branch.

Other context (for non-onboarding PRs)

This updates the existing omniverse-usd-performance-tuning skill in the canonical NVIDIA skills catalog from the merged Physical AI skill hub copy.

This is a patch update for the existing Omniverse manual catalog entry. It brings over the latest USD performance tuning workflow, refreshed validation/reporting references, updated eval metadata, the generated BENCHMARK.md, refreshed skill-card.md, and refreshed skill.oms.sig from the skill hub NVSkills run.

The patch substantially improves the skill's performance workflow by adding stronger Phase-4 target coverage, fail-closed reconciliation behavior, clearer validation-runner guidance, structured validator concepts, and report validation fixtures/scripts.

Validation run locally:

• git diff --check
• JSON parse across skills/omniverse-usd-performance-tuning/**/*.json
• python3 skills/omniverse-usd-performance-tuning/references/optimization-report/scripts/validate_report.py skills/omniverse-usd-performance-tuning/references/report-templates/optimization-report.design-fixture.json --manifest skills/omniverse-usd-performance-tuning/references/report-templates/optimization-report.design-fixture.manifest.json
• python3 skills/omniverse-usd-performance-tuning/references/optimization-report/scripts/validate_report.py skills/omniverse-usd-performance-tuning/references/report-templates/optimization-report.structural-only-fixture.json
• python3 -m py_compile skills/omniverse-usd-performance-tuning/references/optimization-report/scripts/validate_report.py skills/omniverse-usd-performance-tuning/references/usd-validation-runner/scripts/usd_validation_executor.py
• Exact directory comparison against NVIDIA-dev/physical-ai-skill-hub-dev@009a77b:skills/omniverse-usd-performance-tuning

[jensjebens]

/nvskills-ci

[jensjebens]

Update: this branch now includes the generated plugins/nvidia-skills copy required by the catalog plugin drift check. DCO, Verify Authors, and Validate Plugins are green on commit b0a5954.\n\nI attempted /nvskills-ci, but the request workflow failed because it requires the requester to have maintain/admin permission on NVIDIA/skills. A maintainer/admin should re-comment /nvskills-ci to dispatch the NVSkills validation.

[jensjebens]

Thanks for the review. I rebased this branch onto current origin/main (latest head ed276a1) and rechecked the two concerns.\n\nFor item 1: the canonical path now does include the refreshed skill-card.md, skill.oms.sig, references/usd-validation-runner/*, and references/workflow.md. I also verified locally that skills/omniverse-usd-performance-tuning and plugins/nvidia-skills/skills/omniverse-usd-performance-tuning are byte-for-byte identical.\n\nFor item 2: plugins/nvidia-skills is generated by this repo, not copied from the source repo. .github/workflows/validate-plugins.yml runs .github/scripts/build-plugins.sh --check on PRs touching skills/**, and plugins.d/nvidia-skills.yml includes skills/omniverse-usd-performance-tuning/, so omitting the generated plugin copy makes the Validate Plugins check fail with drift. With the generated copy included, DCO, Verify Authors, and Validate Plugins are all green on ed276a1.\n\nThe one remaining external bit is still NVSkills CI: my /nvskills-ci attempt failed because this repo requires maintain/admin permission to dispatch it, so a maintainer/admin needs to re-comment /nvskills-ci.

[jensjebens]

Ran the exact sanity check requested on the current PR head ed276a1:\n\npython3 .github/scripts/build-plugins.py --only nvidia-skills\n\nResult: no working tree changes. I also compared skills/omniverse-usd-performance-tuning against plugins/nvidia-skills/skills/omniverse-usd-performance-tuning file-by-file after the build; counts were missing=0, different=0, extra=0.\n\nGitHub's PR diff for the canonical path includes the refreshed skill-card.md, skill.oms.sig, references/usd-validation-runner/*, and references/workflow.md. No additional push was needed.

[mosheabr]

Approving — apologies for the false alarm in my earlier review. I was looking at gh pr view --json files output which truncated at 100 files (GitHub API default), and the canonical-path updates for skill-card.md, skill.oms.sig, references/usd-validation-runner/*, and references/workflow.md were on the second page so I didn't see them.

Re-verified directly via the contents API:

• Canonical skill-card.md SHA 4671d60 (≠ main 48d29eb) — updated ✓
• Canonical skill.oms.sig SHA bd9e75a (≠ main 5aaecdb) — updated ✓
• Canonical ↔ plugin both at SHA 4671d60 / bd9e75a — byte-for-byte identical ✓
• Validate Plugins CI passing — build-plugins.py --check returns no drift ✓

Good content refresh — Phase-4 target coverage, fail-closed reconciliation, structured validator concepts, the new validate_report.py and usd_validation_executor.py scripts all look solid. Thanks for the patient walk-through on plugins/ being the curated bundle.