qhy991

Metax Report Skill — Development skill for Claude Code

Development community

Profile MetaX (沐曦) GPU kernels with mcTracer / mcProfiler on C550 and C500 — a Claude Code skill.

How to install Metax Report Skill

This entry records only its repository, not the path inside it, so there is no exact command to give. Open qhy991/metax-report-skill and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Metax Report Skill does

Profile MetaX (沐曦) GPU kernels with mcTracer / mcProfiler on C550 and C500 — a Claude Code skill.

Alternatives in Development

Ideal Customer Profile — Identify ICP with demographics, behaviors, and JTBD 7.8k ★
Claude Supermemory — Persistent memory across sessions and projects using Supermemory 2.3k ★
Route Eval — /route-eval — Evaluate a project and set its routing profile 705 ★

README

metax-report-skill

Structured profiling workflow for **MetaX (沐曦) GPU** kernels using the MACA SDK tools **mcTracer** and **mcProfiler**.

Standalone copy of the KerSor plugin's `skills/metax-report-skill/` (KerSor commit b31148ec), with a MetaX C500 section added (`reference/10-c500.md`).

Mirrors the workflow of `ncu-report-skill` and `msprof-report-skill` but targets MetaX C550 / MXC550 (and C500, see `reference/10-c500.md`) with CUDA-compatible source (`mxcc`, `torch.cuda`, Triton MacaDriver).

When to use

Profiling a .cu kernel compiled with mxcc
Reading mcTracer JSON timeline or mcProfiler CSV
Diagnosing memory vs compute bottlenecks on MetaX (dram_pct, sm_pct)
Writing evidence-backed optimization reports for KerSor / AKW metax backend sessions

Quick install (optional global symlink)

git clone https://github.com/qhy991/metax-report-skill
ln -sf "$PWD/metax-report-skill" ~/.claude/skills/metax-report-skill

Layout

Same run-directory convention as `ncu-report-skill`:

profile//
├── REPORT.md
├── harness/
├── reports/          # mcprof.csv, mctrace logs, copied trace JSON
└── analysis/         # helpers output

Helpers

Script	Purpose
`helpers/summarize_profile_run.sh`	Inventory run dir
`helpers/parse_mctrace.py`	mcTracer JSON → kernel latency table
`helpers/analyze_reports.py`	mcProfiler CSV → canonical metrics + A/B
`helpers/metax_utils.py`	Shared parsing library