AGENT SKILL

simready-foundation-add-feature-adapter

Use for adding SimReady feature adapters that mutate USD assets between exact feature or profile versions.

License 许可
Apache-2.0
所属仓库
leemiracle/work4ai
最近更新
2026-11-12
Stars
★ 4
Forks
0

01 前言 Front Matter

name
simready-foundation-add-feature-adapter
license
Apache-2.0
metadata
{ "tags": [ "simready", "adapter", "usd" ], "author": "Shaad Boochoon <sboochoon@nvidia.com>" }
description
Use for adding SimReady feature adapters that mutate USD assets between exact feature or profile versions.

03 SKILL.md 正文

SimReady Add Feature Adapter

Purpose

Use this skill to add a direct asset mutation path between exact FET_###_<RUNTIME> feature names with integer feature versions, or between profile feature bundles. Feature adapters live under nv_core/cip_specs/asset_handler_modules/ and modify an output USD stage so an asset can move from an input feature/profile contract to a target one.

Use this after feature/profile differences are known. If no USD mutation is needed, document that no adapter is required.

Prerequisites

Before editing, read:

  • AGENTS.md
  • nv_core/sr_specs/docs/guides/feature_adapters/feature_adapters.md
  • source and target feature manifests
  • source and target profile versions, if this is profile-driven
  • existing adapters in the relevant asset handler module

Inputs

Collect or infer:

Input Requirement
input_feature_name Source feature name / manifest id, such as FET_003_STANDARD.
input_feature_version Source integer feature version.
output_feature_name Target feature name / manifest id, such as FET_003_PHYSX.
output_feature_version Target integer feature version.
profile_path Optional source/target profile upgrade path.
mutation Exact USD data changes required.
module_path Asset handler module directory.
test_assets Assets that pass input and should pass output after mutation.

Instructions

Use this checklist when changing the repository:

  1. Compare source and target manifests. Confirm names use FET_###_<RUNTIME> and versions are exact integers, then identify only the requirements that differ.
  2. Decide whether a direct adapter is necessary. For profile transformations, every feature difference must have a direct adapter path or a documented blocker. If target requirements are documentation-only or already satisfied by source assets, document no-op behavior.
  3. Choose or create the asset handler module path.
  4. Add a Python adapter file with @feature_adapter metadata using exact feature names and integer versions.
  5. Implement modify_stage(input_stage, output_stage):
    • read source stage only when needed
    • mutate output stage deterministically
    • preserve unrelated authored data
    • save output stage when complete
  6. Use native USD schemas and existing helper patterns from neighboring adapters.
  7. Add focused tests or manual validation notes:
    • input validates against source feature/profile
    • output validates against target feature/profile
    • mutation is idempotent or safely repeatable where practical
  8. Update adapter indexes, imports, or registration files if the repo requires them.
  9. If the adapter is needed for a new profile version, update the profile docs or migration notes.

Examples

Example request:

Add a feature adapter that mutates assets from one exact SimReady feature/profile version to another.

Expected result summary:

changed_files: new docs, manifests, indexes, or validation scaffolding
validation: focused static checks and any relevant docs/build checks
remaining_gaps: requirement, validator, adapter, profile, or runtime-test follow-up

Policies

  • Do not use adapters to hide invalid source assets; adapters should bridge defined feature differences.
  • Do not mutate source stages in place.
  • Do not invent property values that need user intent or prediction unless the adapter contract explicitly defines defaults.
  • Keep profile upgrade paths direct: every changed feature needs a clear adapter path or a documented blocker.

Limitations

  • Do not mutate published feature or profile versions in place.
  • Do not invent requirement IDs or validator behavior when the contract is ambiguous; record the question.
  • Do not skip index, manifest, validation, or downstream follow-up notes.

Troubleshooting

  • Error: the new concept overlaps an existing artifact. Solution: update the existing capability, requirement, feature, profile, or adapter instead.
  • Error: names or IDs conflict. Solution: re-check naming conventions and nearby indexes before editing further.
  • Error: validation strategy is unclear. Solution: document deferred validation and the exact follow-up skill.

Resources

  • assets/openai.yaml preserves optional UI metadata for clients that read skill display hints. It is not required for the workflow.

Summary Format

Report:

Field Meaning
adapter_name Adapter identifier.
adapter_path Python file path.
input_feature Source feature name and integer version.
output_feature Target feature name and integer version.
mutation_summary USD opinions changed.
profile_upgrade Profile path supported, if any.
validation Tests/assets run and remaining gaps.

04 同仓库更多 Skills

leemiracle
leemiracle/work4ai

agentic-ecology-ui

Guidelines and templates for creating a generic web UI to visualize, rank, filter, and annotate vector databases. The UI has a hierarchical structure covering database selection, label selection/definition, ranking/filtering configuration, and row visualization/annotation. Use when building, serving, or adapting an interactive web UI for Hoplite vector databases.

★ 4
leemiracle
leemiracle/work4ai

agentic-ecology-init

Initializes a local uv-managed project directory for agentic ecology workloads. Sets up Python dependencies using reference pyproject.toml and uv.lock, configures workspace rules, and ensures Agentic Ecology skills are discoverable. Use when initializing a new project workspace or bootstrapping a clean environment.

★ 4
leemiracle
leemiracle/work4ai

agentic-ecology-camera-traps

Provides capabilities to run SpeciesNet detector and classifier on camera trap images, extract crop-level feature embeddings, and populate a Hoplite vector database for downstream search and agile modeling. Use when ingesting camera trap images, classifying wildlife photos, or indexing camera trap data.

★ 4
leemiracle
leemiracle/work4ai

agentic-ecology-bioacoustics

Provides bioacoustic analysis capabilities for ecologists and researchers using the perch-hoplite Python package. A typical use case is to use agile modeling to bootstrap the creation and deployment of a bespoke detector for targeted species on an existing collection of passive acoustic monitoring recordings. Use when processing and annotating audio recordings for bioacoustic applications.

★ 4
leemiracle
leemiracle/work4ai

sample-skill

Sample skill used by agent-env integration tests. Exercises SKILL.md parsing plus supporting scripts and references.

★ 4
leemiracle
leemiracle/work4ai

simready-foundation-validate-foundation-change

Use for auditing SimReady requirement, validator, feature, profile, adapter, test, and skill consistency.

★ 4