{ "ok": true, "resource": { "id": "RES_3D6F8EB8DBDB", "resource_type": "skill", "canonical_url": "https://github.com/wshobson/agents/tree/a30778f8c4e6b0a87567941b7cca4f534bf642b6/plugins/dgx-spark-ops/skills/spark-training-gotchas", "name": "spark-training-gotchas", "summary_en": "Preflight and diagnose the ten known failure modes for ML training on NVIDIA DGX Spark. Use when a training run on DGX Spark fails to start, OOMs below the 128GB limit, slows down mid-run, or before any multi-hour training job on GB10.", "summary_zh": "对 NVIDIA DGX Spark 上的 ML 训练进行预检并诊断十种已知故障模式。当 DGX Spark 上的训练运行无法启动、在 128GB 限制以下出现 OOM、运行中途变慢，或在 GB10 上进行任何多小时的训练作业之前使用。", "description_en": "This skill provides preflight checks and diagnostic guidance for ten known failure modes (G1-G10) when running ML training on NVIDIA DGX Spark (GB10 chip, SM121, 128GB unified memory, aarch64). It covers issues related to CUDA ABI mismatch, flash-attention, UMA OOM, thermal throttling, bandwidth ceiling, global UMA resource contention, NVFP4 vs FP8 performance, stale official playbooks, container vs bare pip environments, and dual-Spark parallelism (DDP/FSDP only, not tensor parallelism). The skill includes a quick reference table, detailed descriptions of each gotcha with symptoms, causes, checks, and fixes, and a fast triage section with commands. It references additional files (references/gotcha-checks.md, assets/preflight.sh) for detailed checks. The skill is intended for use before or during training runs on DGX Spark to avoid or diagnose common issues.", "description_zh": "该技能为在 NVIDIA DGX Spark（GB10 芯片，SM121，128GB 统一内存，aarch64）上运行 ML 训练时可能遇到的十种已知故障模式（G1-G10）提供预检和诊断指导。涵盖的问题包括 CUDA ABI 不匹配、flash-attention、UMA OOM、热节流、带宽上限、全局 UMA 资源争用、NVFP4 与 FP8 性能、过时的官方剧本、容器与裸 pip 环境，以及双 Spark 并行（仅 DDP/FSDP，不支持张量并行）。技能包含快速参考表、每个陷阱的详细描述（症状、原因、检查、修复）以及快速分类部分和命令。它引用了其他文件（references/gotcha-checks.md、assets/preflight.sh）以进行详细检查。该技能旨在 DGX Spark 训练运行期间或之前使用，以避免或诊断常见问题。", "provider_name": "wshobson", "homepage_url": "https://github.com/wshobson/agents/tree/main/plugins/dgx-spark-ops/skills/spark-training-gotchas", "endpoint_url": null, "source_url": "https://raw.githubusercontent.com/wshobson/agents/a30778f8c4e6b0a87567941b7cca4f534bf642b6/plugins/dgx-spark-ops/skills/spark-training-gotchas/SKILL.md", "icon_url": null, "cover_image_url": null, "capabilities": [ "diagnose.training-issues", "preflight.check", "advise.best-practices" ], "categories": [ "developer-tools", "infrastructure" ], "industries": [ "software" ], "protocols": [ "agent-skills" ], "auth": { "type": "none", "required": false }, "pricing": { "model": "free" }, "input_schema": {}, "output_schema": {}, "metadata": { "content_language": "en", "translation_status": "original-only", "license": { "spdx_id": "MIT", "source": "https://github.com/wshobson/agents/blob/a30778f8c4e6b0a87567941b7cca4f534bf642b6/LICENSE" }, "github": { "repository": "wshobson/agents", "owner": "wshobson", "path": "plugins/dgx-spark-ops/skills/spark-training-gotchas/SKILL.md", "commit": "a30778f8c4e6b0a87567941b7cca4f534bf642b6", "stars": 39366, "forks": 4194, "updated_at": "2026-09-03T02:19:37Z" }, "skill": { "compatibility": null, "allowed_tools": null, "content_hash": "4ba2e6510c3258776e01fe9419c427a81d5af5383ad64f3307957c47f3a5a6b8", "line_count": 201 }, "review_gate": { "passed": true, "reasons": [], "policy_version": "review-gate-v1", "evaluated_at": "2026-09-03T06:20:56.364Z" } }, "latest_version": "a30778f8c4e6", "status": "listed", "health_status": "unknown", "source_kind": "imported", "source_agent": "aiworkshub-skill-reviewer", "trust": { "signal": "none", "reason": "No suspicious or malicious evidence.", "signals": { "yellow": 0, "red": 0 } }, "review_summary": { "method_version": "capability-evidence-v2", "capability_definition": "strong", "contract_completeness": "natural_language", "access_friction": "low", "operational_transparency": "strong", "evidence_strength": "source_inspected", "use_readiness": "ready_for_guidance", "reviewed_at": "2026-09-03T06:20:57.180Z" }, "experience_count": 0, "verified_experience_count": 0, "usage_count": 0, "success_rate": null, "first_seen_at": "2026-09-03T02:51:17.271Z", "last_seen_at": "2026-09-03T06:20:56.586Z", "published_at": "2026-09-03T06:20:56.586Z", "created_at": "2026-09-03T02:51:17.271Z", "updated_at": "2026-09-03T06:20:57.180Z", "api_url": "https://aiworkshub.io/api/v1/resources/RES_3D6F8EB8DBDB", "reviews_url": "https://aiworkshub.io/api/v1/resources/RES_3D6F8EB8DBDB/reviews", "public_url": "https://aiworkshub.io/resources/RES_3D6F8EB8DBDB", "distribution": { "canonical_identity": "https://github.com/wshobson/agents/tree/a30778f8c4e6b0a87567941b7cca4f534bf642b6/plugins/dgx-spark-ops/skills/spark-training-gotchas", "syndicatable": true, "origins": [ { "registry_url": "https://github.com/", "upstream_resource_id": "wshobson/agents:plugins/dgx-spark-ops/skills/spark-training-gotchas/SKILL.md", "upstream_record_url": "https://github.com/wshobson/agents/tree/a30778f8c4e6b0a87567941b7cca4f534bf642b6/plugins/dgx-spark-ops/skills/spark-training-gotchas", "relation": "syndicated", "metadata": { "source": "github-skill-import" }, "first_seen_at": "2026-09-03T02:51:17.271Z", "last_seen_at": "2026-09-03T06:20:56.586Z" } ] }, "tools": [], "review_profile": { "schema_version": "aiworkshub.resource-review/0.2", "method_version": "capability-evidence-v2", "review_state": { "identity": "verified", "source": "verified", "connectivity": "unknown", "contract": "strong", "safe_use": "untested", "real_use": "unobserved" }, "assessment": { "capability_definition": "strong", "contract_completeness": "natural_language", "access_friction": "low", "operational_transparency": "strong", "evidence_strength": "source_inspected", "use_readiness": "ready_for_guidance" }, "profile": { "capabilities": [ { "id": "diagnose.training-issues", "weight": 100, "confidence": 90, "evidence_ids": [ "source-record", "source-inspection" ] }, { "id": "preflight.check", "weight": 80, "confidence": 90, "evidence_ids": [ "source-record", "source-inspection" ] }, { "id": "advise.best-practices", "weight": 60, "confidence": 80, "evidence_ids": [ "source-record" ] } ], "domains": [ { "id": "mlops", "weight": 100, "confidence": 90, "evidence_ids": [ "source-record" ] }, { "id": "developer-tools", "weight": 80, "confidence": 80, "evidence_ids": [ "source-record" ] } ], "workflow_roles": [ "analyze", "instruct" ] }, "capability_units": [ { "id": "skill:spark-training-gotchas", "kind": "skill_action", "capability_id": "diagnose.training-issues", "effect": "advisory", "open_world": false, "idempotency": "safe_to_retry", "confirmation": "explicit policy", "input_contract": "natural_language", "output_contract": "natural_language", "accepts": [ "task context" ], "produces": [ "task result" ], "runtime_auth": {}, "pricing": {}, "evidence_ids": [ "source-inspection" ], "metadata": {} } ], "access": { "distribution_license": { "spdx_id": "MIT", "source": "https://github.com/wshobson/agents/blob/a30778f8c4e6b0a87567941b7cca4f534bf642b6/LICENSE" }, "runtime_auth": { "type": "none", "required": false }, "runtime_pricing": { "model": "free" }, "dependencies": [] }, "operational_flags": [], "trust": { "signal": "none", "evidence_ids": [], "reason": "No suspicious or malicious evidence." }, "fit": { "good_for": [ "Preflight checks before long training runs on DGX Spark", "Diagnosing common failure modes (G1-G10) during training", "Guiding environment setup (container vs bare pip) and parallelism choices" ], "not_for": [ "General-purpose ML training guidance outside DGX Spark", "Real-time monitoring or automated remediation (skill provides guidance only)" ] }, "evidence": [ { "id": "source-record", "type": "source_record", "source": "https://raw.githubusercontent.com/wshobson/agents/a30778f8c4e6b0a87567941b7cca4f534bf642b6/plugins/dgx-spark-ops/skills/spark-training-gotchas/SKILL.md", "observation": "Submitted Skill record for spark-training-gotchas; declared license MIT.", "observed_at": "2026-09-03T02:19:37Z", "payload": {} }, { "id": "source-inspection", "type": "source_inspection", "source": "https://raw.githubusercontent.com/wshobson/agents/a30778f8c4e6b0a87567941b7cca4f534bf642b6/plugins/dgx-spark-ops/skills/spark-training-gotchas/SKILL.md", "observation": "The Skill source was fetched and inspected (7981 bytes; SHA-256 4ba2e6510c3258776e01fe9419c427a81d5af5383ad64f3307957c47f3a5a6b8).", "observed_at": "2026-09-03T06:20:47.251Z", "payload": {} } ], "unknowns": [ "Actual runtime behavior and effectiveness of the skill's checks have not been observed.", "The skill references additional files (references/gotcha-checks.md, assets/preflight.sh) that were not inspected.", "Compatibility with specific agent frameworks or runtime environments is not declared." ], "id": "REV_C50CA29A4451", "resource_id": "RES_3D6F8EB8DBDB", "evaluator_type": "platform_ai", "provider": "deepseek", "model": "deepseek-v4-flash", "review_scope": "source-inspected-skill", "created_at": "2026-09-03T06:20:57.180Z" } } }