| name | asm-analysis |
| description | Use when the user asks to look at, analyse, profile, or optimise
ASSEMBLY or the runtime performance of compiled C/C++ — from a source
file, CMake/Makefile target, object file, binary, or raw .s listing.
TRIGGER on: "look at the asm", "inspect codegen", "why is this loop
slow", "can this be vectorised", "what intrinsic replaces this",
"annotate this .s", and on any "make this faster / vectorise / tune /
maximise throughput" request for C/C++, including hardware-specific
ones ("faster on Zen4", "tune for Skylake-X", "use AVX-512", "avoid
the port-5 bottleneck"). Default here whenever the goal is runtime
performance of compiled C/C++, unless scoped to source-only
(readability, algorithmic complexity, API ergonomics).
DO NOT TRIGGER for pure source refactoring, style, or C++ semantics
questions that don't require reading generated code. |
asm-analysis
Entrypoint for the shared simdref assembly analysis workflow.
Before doing any assembly analysis, read and follow
references/workflow.md. Treat that file as the source of truth for the
compile → objdump/-S → simdref annotate → simdref llm batch → proposal
pipeline.
Claude Code install and updates
This skill lives at skills/asm-analysis/SKILL.md inside the simdref
source tree, which is also published as a Claude Code plugin
marketplace. Preferred install at the Claude Code prompt:
/plugin marketplace add DiamonDinoia/simdref
/plugin install asm-analysis@simdref
To refresh later:
/plugin marketplace update simdref
/plugin install asm-analysis@simdref
Manual alternatives, either symlink from a checkout for always-current
updates or install a one-off snapshot:
git clone https://github.com/DiamonDinoia/simdref.git ~/src/simdref
mkdir -p ~/.claude/skills
ln -sf ~/src/simdref/skills/asm-analysis ~/.claude/skills/asm-analysis
mkdir -p ~/.claude/skills/asm-analysis && \
curl -fsSL https://raw.githubusercontent.com/DiamonDinoia/simdref/main/skills/asm-analysis/SKILL.md \
-o ~/.claude/skills/asm-analysis/SKILL.md
If the user has an older PyPI-shipped copy, point them at the packaged
location for the upgrade diff:
pip show -f simdref | grep SKILL.md