NVIDIA/TensorRT-Model-Connect is a collection of C++ reference implementations for deploying supported Hugging Face models with NVIDIA TensorRT, a system that runs trained models to produce inference results. It is for developers who want to build and run supported models or evaluate model integrations through TensorRT. The catalogue entries provide skills and instructions for using this model deployment workflow.
Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/NVIDIA/TensorRT-Model-Connectnpx agentmods add skills/nvidia/tensorrt-model-connect/review-trtmc-prWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nvidia/tensorrt-model-connect/review-trtmc-pr)<a href="https://agentmods.dev/skills/nvidia/tensorrt-model-connect/review-trtmc-pr"><img src="https://agentmods.dev/badge/skills/nvidia/tensorrt-model-connect/review-trtmc-pr/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nvidia/tensorrt-model-connect/review-trtmc-pr"><img src="https://agentmods.dev/badge/skills/nvidia/tensorrt-model-connect/review-trtmc-pr.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00105 | $0.03981 |
| Opus 5.5 | $0.00042 | $0.01592 |
| Sonnet 5.5 | $0.00021 | $0.00796 |
| Haiku 4.5 | $0.00011 | $0.00398 |
Grade A, and why
review-trtmc-pr 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 8d 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.
How it starts
The opening of the file, as written. The whole thing — 362 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Review TRTMC PR
Review Contract
- Review an explicit comparison: the exact PR base and head in pull-request
mode, or the fetched canonical
main, localHEAD, and disclosed working tree state in local self-review mode. Never substitute a remembered branch. - Treat the canonical
NVIDIA/TensorRT-Model-Connect:main, rootAGENTS.md,CONTRIBUTING.md, and current architecture documents as authoritative. PR #1093 explains the cutover but is historical rationale, not a substitute for current code and policy. - Review changed behavior and newly relied-on behavior. Do not charge the PR for unrelated, unchanged migration debt.
- Keep the PR and contributor branch unchanged. Do not approve, request changes, comment, label, trigger CI, push, or edit code unless the user explicitly asks for that separate action.
- Treat fork code as untrusted. Inspect it in a detached temporary worktree; never run it on a privileged or protected runner and never expose secrets.
- An absence of findings is not proof of correctness. State untested paths and unresolved evidence explicitly.
Select The Review Mode
- Use pull-request mode when given a PR number or URL. Prefer this mode for the final contributor self-review because it includes the exact remote head, PR description, commits, linked issue, and current checks.
- Use local self-review mode when asked to review a branch before a PR exists. Include committed changes and disclose staged, unstaged, and untracked files. A dirty working tree cannot receive a ready-to-submit result because its uncommitted content has no immutable reviewed head.
For contributor self-review, remain read-only and tell the contributor whether the self-review checkbox can honestly be selected. Do not mark the PR ready, push, commit, edit files, or publish comments unless separately requested.
Establish The Pull-Request Baseline
Record the repository, PR number, base SHA, head SHA, author, linked issue, current review state, and check state. Refresh remote information instead of using a prior session or stale local ref.
What ships with it
3 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.
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.
- 8d ago Changed · +10 lines · +7 tokens per session 3464842e5e4b
- 1mo ago Changed a9633e6227f6
- 1mo ago First seen · 352 lines · 98 tokens per session scan A f0afbe91e66a
review-trtmc-pr is a skill published in the GitHub repository NVIDIA/TensorRT-Model-Connect (280 stars, last pushed yesterday), licensed Apache-2.0. It adds 105 tokens to every session and 3,981 once invoked, about $0.0004 per session on Opus 5.5. 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 2026-09-08.
Other skills, from other repositories
huggingface-transformers
Use when running or fine-tuning Hugging Face Transformers models in Python — pipeline() inference for text classification and generation, loading weights with AutoModel/AutoTokenizer, devicemap and dtype memory management, tokenizer usage, Trainer fine-tuning, and BGE embeddings for Chinese RAG. Covers transformers…
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.
hwdef-check
Review an ArduPilot new-board hwdef PR. Stashes any local changes, checks the PR out as a branch in the current repo, runs DMA / board-ID / file-presence / commit-structure checks, layers the hwdef playbook on top, and (after confirmation) posts a review comment on the PR — then restores the original branch and pops…
change-review
Interactive AI skill (Claude Code or Codex) for reviewing dispatch-completed changes. Uses branch-centric review, performs hard checks and three-dimension Verify before merge, then publishes specs/archive/backlog updates through a recoverable governance PR. Use when dispatch runner has completed task groups (status…
pr
Draft a PR body from OMC's paper trail — the smallest visual that proves the change, evidence consumed from the verify protocol (never re-collected), reversibility from the ADR test on the diff, glossary language throughout.
otto-pr-review-loop
Run Polly and OCR on a PR until all findings settle.