> For clean Markdown content of this page, append .md to this URL. For the complete documentation index, see https://docs.nvidia.com/dynamo/llms.txt. For full content including API reference and SDK examples, see https://docs.nvidia.com/dynamo/llms-full.txt.

# Agent Skills

Dynamo includes agent skills for development, documentation, deployment, operations, and
performance optimization. Compatible coding agents discover the skills from the repository and
select them based on your request. Describe the outcome you want; you do not normally need to name
a skill explicitly.

The canonical sources live in
[`.agents/skills/`](https://github.com/ai-dynamo/dynamo/tree/v1.5.0/.agents/skills). The `skills/` and
`.claude/skills/` directories are compatibility symlinks to the same content.

#### [Deployment and Operations](/dynamo/agent-skills/deployment)

Deploy a recipe, start a router, validate interconnects, and diagnose unhealthy deployments.

#### [Performance Optimization](/dynamo/agent-skills/optimization)

Run the evidence-driven loop that benchmarks a baseline and challenges it with candidates.

#### [Development and Documentation](/dynamo/agent-skills/development)

Investigate bugs, file enhancement proposals, review Rust changes, audit CI, and author docs.

## Example Prompts

Clone the repository, point your agent at it, and type a sentence like one of these. The agent
picks the skill; you state the outcome, the constraints, and the budget.

* Deploy the Qwen3 32B FP8 recipe on my cluster and smoke-test the endpoint.
* Optimize this deployment for output tokens per second per user under a 200 ms time-to-first-token
  Service Level Objective (SLO). Budget 8 GPU-hours and stop after three failed deployments.
* My prefill pods are restarting after a rolling update. Find out why.
* CI failed on my pull request. Tell me whether it is a real regression or a flake.
* Draft a Dynamo Enhancement Proposal for the change I just described, then show it to me before
  filing.

## Usage Notes

* **Invoke by intent.** Ask for the result you need, such as deploying a model, investigating a
  regression, reviewing a pull request, or optimizing against a Service Level Objective (SLO).
* **Set budgets for optimization.** State GPU-hour, wall-clock, and failed-deployment limits before
  starting a long-running optimization loop.
* **Use goal mode for unattended runs.** Answer the agent's workload questions before enabling the
  harness's goal or budget mode.
* **Protect credentials.** Do not provide secrets in prompts or allow credentials into generated run
  artifacts.