技能创建器
skill-creator
创建、改进和评估自定义技能,把专业流程转化为可重复使用的 AI 能力。
behavior-1k【数据与研究】CPU-only operating guidance for the BEHAVIOR-1K BDDL 3.7.0 package: symbolic activity definitions, condition parsing/evaluation, object taxonomy, and generated-data-backed knowledge-base inspection.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this skill when a Researcher needs the BDDL portion of BEHAVIOR-1K: activity-definition lookup, parsing, symbolic condition reasoning, object synset/category taxonomy, or read-only knowledge-base inspection.
.bddl files, predicates, initial/goal conditions, scope,
grounding, or symbolic evaluation: read
symbolic-tasks.KnowledgeBase models: read
knowledge-base.This generated skill intentionally covers only CPU-verifiable BDDL package workflows from the pinned repository snapshot. Install the public distribution in a compatible Python environment:
python -m pip install "bddl==3.7.0"
python -c "from importlib.metadata import version; import bddl; print(version('bddl'))"
The package includes activity/domain definitions and generated taxonomy data as package-relative runtime data. Use the bundled read-only inspectors for quick checks, then consult the sub-skill API references for programmatic workflows:
python sub-skills/symbolic-tasks/scripts/inspect_bddl_activity.py --help
python sub-skills/knowledge-base/scripts/inspect_bddl_taxonomy.py --help
python scripts/smoke_bddl_install.py --help
This skill does not cover OmniGibson, Isaac Sim, simulator/environment execution, object-state runtime behavior, robots, controllers, sensors, action primitives, cuRobo, GPU backends, physical hardware, asset production, benchmark orchestration, or the excluded BEHAVIOR-1K monorepo packages. A symbolic BDDL condition is not evidence that a simulator can instantiate or satisfy that condition. The source snapshot's OmniGibson/Isaac Sim workflows were not generated or verified because the required runtime was unavailable; do not route such requests here as if CPU guidance were a substitute.
The generated-data files are runtime inputs. Do not run maintainer-only
data_generation pipelines as a routine repair or query method. Use the
troubleshooting references when package data is missing or inconsistent.
For a quick, read-only package/data smoke check, run
scripts/smoke_bddl_install.py. For activity-level and taxonomy-level
validation, use the inspectors linked above. Keep all simulator-dependent
claims outside this skill.
See provenance for the source revision and evidence overview for the retained package surface.