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.
npx agentmods add skills/nvidia/tensorrt-model-connect/transform-modelnpx skills add NVIDIA/TensorRT-Model-Connect --skill transform-modelgit clone --depth 1 https://github.com/NVIDIA/TensorRT-Model-ConnectWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00063 | $0.01640 |
| Opus 5 | $0.00032 | $0.00820 |
| Sonnet 5 | $0.00013 | $0.00328 |
| Haiku 4.5 | $0.00006 | $0.00164 |
Grade A, and why
transform-model 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 yesterday.
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.
How it starts
The opening of the file, as written. The whole thing — 197 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Transform Model
Define The Support Claim
Record:
- exact Hugging Face model ID and immutable revision;
- task/modality and requested public operation;
- target hardware and precision/quantization;
- closest existing family and architectural differences;
- whether the expected bundle path is native or an exact optimized profile;
- requested evidence level: build, parity, E2E, performance, or qualification.
Do not begin from a generic manifest. Read the model config, reference implementation, nearest family descriptors, and owned tests first.
Ownership Map
A fully registered native model normally crosses three descriptors:
| Layer | Owner |
|---|---|
| Python build/family selection | python/tensorrt_model_connect/families/<family>/MODEL.toml |
| Native C++ strategy/plugin | src/runtime/models/<family>/MODEL.toml |
| E2E models, manifests, sidecars | tests/e2e/models/<family>/MODEL.toml |
The owning directories also contain family-local builders, graph helpers, runtime sources, manifests, testcases, thresholds, and performance contracts. Keep changes there unless multiple families demonstrably share the contract.
An optimized-runtime path instead requires an exact implementation, profile, and qualification chain. Do not create a native strategy merely to mirror an optimized implementation, and do not silently fall back after an optimized profile has claimed the request.
Choose Reuse Or A New Family
Extend an existing family when model type, checkpoint mapping, graph dataflow, runtime strategy, and validation contract fit that owner. Create a new family when those contracts materially differ.
For a standard decoder starting point:
python3 scripts/new_family.py \
--model-type <model-type> \
--hf-repo <org/model> \
--family-name <family>
The scaffold is only a starting point. It fetches model configuration and emits a decoder-oriented Python family. Review every generated match rule, import, weight mapping, graph path, and descriptor; do not use it for non-decoder architectures without redesigning the generated code.
What ships with it
1 file 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.
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.
- yesterday First seen · 197 lines · 63 tokens per session scan A 8d2666374edd
transform-model is a skill published in the GitHub repository NVIDIA/TensorRT-Model-Connect (188 stars, last pushed yesterday), licensed Apache-2.0. It adds 63 tokens to every session and 1,640 once invoked, about $0.0003 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.
Other skills, from other repositories
graphsignal-profiler
Set up the Graphsignal Profiler for inference workloads — vLLM, SGLang, PyTorch, and dstack services. Use when the user wants GPU profiling, tracing, or monitoring for inference, asks about graphsignal-run or graphsignal.watch(), or asks about CUPTI / Prometheus / OTLP setup.
weights-and-biases
Track ML experiments with automatic logging, visualize training in real-time, optimize hyperparameters with sweeps, and manage model registry with W&B - collaborative MLOps platform.
huggingface-accelerate
Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.
mlflow
Track ML experiments, manage model registry with versioning, deploy models to production, and reproduce experiments with MLflow - framework-agnostic ML lifecycle platform.
perf-host-analysis
Analyze host/CPU overhead in TensorRT-LLM inference from nsys traces. Detect whether host overhead is the bottleneck using GPU idle ratio, host prep exposed ratio, and per-phase evidence. For regressions, isolate forward steps via allreduce/NVTX patterns, compare host operation breakdowns across versions, and identify…
perf-host-optimization
Profiles and optimizes TensorRT-LLM host/CPU overhead using lineprofiler (with nsys support planned). Runs iterative profile-analyze-optimize-validate rounds. Use when GPU utilization is low or optimizing PyExecutor throughput.