agentmods.dev

perf-optimization-casebook

A collection of documented TensorRT-LLM performance-optimization examples and decision precedents. TensorRT-LLM is a system for running large language models with NVIDIA GPUs.

In plain words
What is it for?
Use it to find prior solutions for runtime, execution, or GPU-kernel bottlenecks and adapt them to a particular model, configuration, or hardware setup.
Why use it?
It helps compare a current performance problem with approaches that have worked before instead of choosing an optimization without relevant examples.

Skill for Claude CodeCodex

15k 3d ago A scan 91 tokens no licence
View the source file NVIDIA/TensorRT-LLM
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/nvidia/tensorrt-llm/perf-optimization-casebook
Any agent
npx skills add NVIDIA/TensorRT-LLM --skill perf-optimization-casebook
Clone the repo
git clone --depth 1 https://github.com/NVIDIA/TensorRT-LLM

Made for: Claude Code, Codex.

Per session 91 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,435 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin unknown No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00091 $0.04435
Opus 5 $0.00046 $0.02218
Sonnet 5 $0.00018 $0.00887
Haiku 4.5 $0.00009 $0.00443

Measured 2d ago against content hash c28fef69a23d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

perf-optimization-casebook scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/perf-optimization-casebook/SKILL.md · 294 lines

The source is not reproduced here

No licence file

A repository with no LICENSE is all rights reserved by default, so the body is not copied here. The metadata, the measurements and the link are.

Read it on GitHub

Files

What ships with it

49 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 294 lines · 91 tokens per session scan A c28fef69a23d

Subscribe to this mod's changes

perf-optimization-casebook is a skill published in the GitHub repository NVIDIA/TensorRT-LLM (14,505 stars, last pushed 3d ago), with no licence file. It adds 91 tokens to every session and 4,435 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Data and AIOptimizationPyTorchblackwellcasebookcudadecision-supportllm-servingmoeprecedents
Related

Other skills, from other repositories

sglang-diffusion-performance

Use when choosing the fastest SGLang Diffusion flags for a model, GPU, and VRAM budget.

sgl-project/sglang · 29 tokens

add-jit-kernel

Step-by-step tutorial for adding a new lightweight JIT CUDA kernel to sglang's jitkernel module.

sgl-project/sglang · 27 tokens

sglang-runtime-context

How SGLang's runtime configuration and process-global state are organized (RuntimeContext tiers, publish + namespace config bags, the pristine ServerArgs seed, override entry points, resource/stream/buffer leases, per-forward flags), the CI guardrails that enforce the design, and the idioms for developing and testing…

sgl-project/sglang · 93 tokens

sglang-diffusion-add-model

Use when adding a new diffusion model or Diffusers pipeline to SGLang.

sgl-project/sglang · 24 tokens

llm-torch-profiler-analysis

Unified LLM torch-profiler triage skill for sglang, vllm, TensorRT-LLM, and TokenSpeed. Use it to inspect an existing trace.json(.gz) or profile directory, or to drive live profiling against a running server when supported and return one three-table report with kernel, overlap-opportunity, and fuse-pattern tables.

sgl-project/sglang · 84 tokens

ci-workflow-guide

Guide to SGLang CI workflow orchestration — stage ordering, fast-fail, gating, partitioning, execution modes, and debugging CI failures. Use when modifying CI workflows, adding stages, debugging CI pipeline issues, or understanding how tests are dispatched and gated across stages.

sgl-project/sglang · 60 tokens