From amd-skills
Orchestrates modular PyTorch profiler trace analysis with TraceLens: generates perf reports, prepares category data, runs system-level and compute-kernel subagents in parallel, validates outputs, and writes a prioritized stakeholder report (analysis.md). Use when the user asks to follow the analysis orchestrator, run the agentic analysis workflow, analyze a trace, compare two traces, or mentions standalone or comparative TraceLens analysis.
How this skill is triggered — by the user, by Claude, or both
Slash command
/amd-skills:tracelens-analysis-orchestratorThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
<!--
agents/convolution-analyzer.mdagents/cpu-idle-analyzer.mdagents/elementwise-analyzer.mdagents/gemm-analyzer.mdagents/generic-op-analyzer.mdagents/kernel-fusion-analyzer.mdagents/model-identification-agent.mdagents/moe-analyzer.mdagents/multi-kernel-analyzer.mdagents/norm-analyzer.mdagents/reduce-analyzer.mdagents/sdpa-analyzer.mdagents/triton-analyzer.mdevals/evals.pyreference.mdskill-card.mdtemplates/analysis_template.mdtemplates/sub_agent_spec.mdCoordinate system-level analysis (CPU/idle, kernel fusion, multi-kernel / comm / memcpy) and compute-kernel analysis (GEMM, SDPA, elementwise, etc.): one trace load, shared prep, parallel subagents, then aggregation into analysis.md.
Follow reference.md for every step (user prompts, <prefix> / {CMD} usage, CLI commands, subagent launch text, validation, report tee order, plot embedding, and trace diagnostics).
0. Query User Inputs (Platform, Trace Path(s), Analysis Mode, Environment Setup)
1. Generate Performance Report (branches on analysis mode: training vs inference then, comparison scope)
2-5. Prepare Category Data (GPU Util, Top Ops, Tree Data, Multi-Kernel Data, Category Filtering)
6. System-Level Analysis (PARALLEL) → system_findings/
7. Compute Kernel Subagents (PARALLEL) → category_findings/
7.5. Aggregate → priority_data.json::findings[]
8. Validate Subagent Outputs
9. load_findings + Model Identification (subagent) → metadata/model_info.json
10. Render performance PNG if agent_extension.py is absent
11. Generate analysis.md (orchestrator writes via <prefix> tee), optional extension, embed PNG
<output_dir>/cache/cmd_prefix.txt ({CMD} substitution).TraceLens/Agent/Analysis/skills/analysis-orchestrator/agents/<name>.md (see reference.md for exact paths and prompt shells).<output_dir>/analysis.mdsystem_findings/, category_findings/, category_data/, metadata/, perf_report*.xlsx, CSV folders — see package README for layout.Project subagents ship with this skill: TraceLens/Agent/Analysis/skills/analysis-orchestrator/agents/*.md.
claude plugin install amd-skills@claude-plugins-officialAnalyzes torch.profiler traces from sglang, vLLM, TensorRT-LLM, and TokenSpeed. Provides triage with kernel, overlap-opportunity, and fuse-pattern tables.
Analyzes Huawei Ascend NPU profiling data to detect performance anomalies and reverse-engineer model architecture. Outputs a Markdown report with bubble detection, wait-anchor analysis, and layer classification.
Analyzes NVIDIA Nsight Systems GPU profiles (.sqlite/.nsys-rep) to identify root causes and fixes for bottlenecks, NCCL slowdowns, and efficiency issues.