tensorrt-llm

Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency. Use for production deployment on NVIDIA GPUs (A100/H100), when you need 10-100x faster inference than PyTorch, or for serving models with quantization (FP8/INT4), in-flight batching, and multi-GPU scaling.

Quality

86%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

72%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

An actionable, well-structured skill body with executable examples and properly signaled references that resolve to real files. Its main weaknesses are version pins embedded in the main flow (staleness risk) and a deployment workflow lacking validation checkpoints.

Suggestions

Move hard version pins (tensorrt_llm==1.2.0rc3, CUDA 13.0.0, TensorRT 10.13.2) into a version-matrix or 'compatibility' section so the quick-start does not go stale.

Add a validation/verification checkpoint to the serving workflow (e.g., a health check curl or expected startup log line) before claiming the server is ready.

Trim marketing-style benchmark claims ('100× faster') or relocate detailed numbers to a reference file so the overview stays token-efficient.

Dimension	Reasoning	Score
Conciseness	The body is mostly lean but embeds hard version pins ('tensorrt_llm==1.2.0rc3', 'CUDA 13.0.0, TensorRT 10.13.2') in the main quick-start flow rather than a deprecated section, and includes marketing-style benchmark claims ('100× faster') that add staleness risk.	2 / 3
Actionability	It provides fully executable, copy-paste-ready code and shell commands across basic inference, FP8 quantization, multi-GPU deployment, and serving with curl.	3 / 3
Workflow Clarity	The install → configure → serve steps are sequenced, but there are no explicit validation checkpoints or feedback loops for the multi-step deployment workflow.	2 / 3
Progressive Disclosure	The body points to three clearly signaled, one-level-deep references (optimization.md, multi-gpu.md, serving.md) that all exist on disk, keeping the overview lean while splitting detail out.	3 / 3
	Total	10 / 12 Passed

Description

100%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, specific description that states concrete capabilities, includes natural trigger terms, and provides explicit 'Use for' guidance. It uses third-person voice and reads as a clear, non-generic niche.

Dimension	Reasoning	Score
Specificity	The description lists multiple concrete capabilities — 'quantization (FP8/INT4)', 'in-flight batching', and 'multi-GPU scaling' — rather than naming only the domain.	3 / 3
Completeness	It answers both 'what' (Optimizes LLM inference with NVIDIA TensorRT) and 'when' via an explicit 'Use for...' clause with concrete triggers.	3 / 3
Trigger Term Quality	It surfaces natural user terms such as 'LLM inference', 'NVIDIA TensorRT', 'production deployment on NVIDIA GPUs (A100/H100)', 'FP8/INT4', and 'multi-GPU scaling' that a user would actually say.	3 / 3
Distinctiveness Conflict Risk	The NVIDIA TensorRT / FP8 / A100-H100 niche is distinct and unlikely to trigger for unrelated skills.	3 / 3
	Total	12 / 12 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: OpenLAIR/dr-claw
Path: skills/inference-serving/tensorrt-llm/SKILL.md
Commit: 6da7f7c

Reviewed: about 7 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.