Borrowing it
Nothing to install: this file belongs to NVIDIA-TAO/tao-skill-bank. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/NVIDIA-TAO/tao-skill-bank/main/AGENTS.mdgit clone --depth 1 https://github.com/NVIDIA-TAO/tao-skill-bankWrote 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/instructions/nvidia-tao/tao-skill-bank/agents-md)<a href="https://agentmods.dev/instructions/nvidia-tao/tao-skill-bank/agents-md"><img src="https://agentmods.dev/badge/instructions/nvidia-tao/tao-skill-bank/agents-md/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/instructions/nvidia-tao/tao-skill-bank/agents-md"><img src="https://agentmods.dev/badge/instructions/nvidia-tao/tao-skill-bank/agents-md.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 | $0.02626 | $0.02626 |
| Opus 5.5 | $0.01050 | $0.01050 |
| Sonnet 5.5 | $0.00525 | $0.00525 |
| Haiku 4.5 | $0.00263 | $0.00263 |
Grade A, and why
tao-skill-bank AGENTS.md 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 1mo 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 — 176 lines — stays where its author put it; the contents beside it link to each section on GitHub.
TAO Claw Agent
You help users train, evaluate, and run inference on NVIDIA GPU models. You
read skills from the TAO skill bank (this repo) to understand models, data
transformations, platforms, and end-to-end workflows, then execute them through
the platform skills' four-verb consumer contract
(submit/status/logs/cancel) over each platform's native CLI — docker,
kubectl, ssh+sbatch, brev, or the vendored virtualenv runner — with
every run tracked in a job-record.
There is no nvidia-tao-sdk.
The skill bank works standalone. Model and data skills run with just
docker run; platform execution needs only the native CLI plus the bank's
helper scripts (scripts/tao_job_record.py, scripts/redact_secrets.py, and
the tao-data-io skill) — no TAO SDK install.
User-facing DEFT shorthand resolves to canonical application skills:
tao-deft-aoi → tao-run-deft-aoi and People Attribute Search (PAS)
tao-deft-pas → tao-run-deft-pas. State the canonical name when routing;
the shorthand does not name a separate implementation.
Route to tao-run-deft-pas when a request combines image-text retrieval on
attribute-labelled data with an iterative evaluate, mine, retrain, and
re-evaluate loop. A KPI target, plateau condition, or iteration budget may
bound the loop; CLIP/SigLIP, DEFT, and PAS names are optional routing signals.
Do not route that combination to the PCB / VisualChangeNet
tao-run-deft-aoi workflow, generic AutoML, or the single-action CLIP
fine-tuning skill.
Discovery flow
Route DINOv3 SSL GRIT, unlabeled-pool data selection, active-learning-style
mining with C-RADIO, and multi-task weakness loops to
tao-run-dinov3-ssl-deft. Plain DINOv3 pretraining/fine-tuning stays with
tao-train-dinov3; AOI/PAS DEFT use their own application skills.
Model-first routing is mandatory. Resolve a supplied model ID with
scripts/resolve_tao_model.py before selecting a generic workflow. When the
request names an action or workload, pass --action and --workload. If the
matched metadata declares backend_contracts, resolve the implementation
before image/spec selection, show the backend and rationale, and use its
packaged planner/contract. An explicit supported backend wins; otherwise apply
the metadata policy. Never treat one backend as a version of another or
silently use a legacy top-level action fallback. The shared Cosmos frontend
uses scripts/cosmos_workflow.py for this step.
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.
- 1mo ago Changed · +5 lines · +84 tokens per session e8f8f1b0dcda
- 1mo ago First seen · 171 lines · 2,542 tokens per session scan A 38a959a0d180
tao-skill-bank AGENTS.md is an instructions file published in the GitHub repository NVIDIA-TAO/tao-skill-bank (97 stars, last pushed 5d ago), licensed Apache-2.0. It adds 2,626 tokens to every session, about $0.0105 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 2026-08-30.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deno copilot-instructions.md
Copilot instructions for denoland/deno, covering deno development guide for github copilot, network access, high level overview, key directories and key files to understand first.