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 skills add llm-d-incubation/llm-d-skills --skill llm-d-autoconfiggit clone --depth 1 https://github.com/llm-d-incubation/llm-d-skillsWrote 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/llm-d-incubation/llm-d-skills/llm-d-autoconfig)<a href="https://agentmods.dev/skills/llm-d-incubation/llm-d-skills/llm-d-autoconfig"><img src="https://agentmods.dev/badge/skills/llm-d-incubation/llm-d-skills/llm-d-autoconfig/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/llm-d-incubation/llm-d-skills/llm-d-autoconfig"><img src="https://agentmods.dev/badge/skills/llm-d-incubation/llm-d-skills/llm-d-autoconfig.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 | $0.00075 | $0.03548 |
| Opus 5 | $0.00037 | $0.01774 |
| Sonnet 5 | $0.00015 | $0.00710 |
| Haiku 4.5 | $0.00007 | $0.00355 |
Grade A, and why
llm-d-autoconfig scanned grade A with 1 finding 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 12d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
| 1 | Cluster discovery (read-only kubectl/curl) | [references/phase-1-cluster-discovery.md](references/phase-1-cluster-discovery.md) | How it starts
The opening of the file, as written. The whole thing — 110 lines — stays where its author put it; the contents beside it link to each section on GitHub.
llm-d-autoconfig
You help the user configure (and optionally deploy) an EndpointPickerConfig for their llm-d inference deployment.
This skill wraps a deterministic recommender script (autoconfig). Your job is to gather workload information conversationally, hand it to the script in a structured form, narrate the output with its evidence tiers, and — if the user wants — apply the result to their cluster.
Relationship to the single-purpose skills
This repo also ships focused skills — deploy-llm-d, run-llm-d-benchmark, compare-llm-d-configurations, configure-wva-autoscaling-llm-d, teardown-llm-d. This skill owns config generation and the clone-free deterministic bundle; the single-purpose skills own standalone operations on a config the user already has.
- Stay in this skill when the user is deciding what config to run for a workload/SLA, then deploying/benchmarking the bundle it renders. Phases 6–7 run the bundle-native, URL-only (no repo clone) deploy + deterministic benchmark — that path is autoconfig's and is not duplicated elsewhere.
- Hand off to a single-purpose skill when the user's ask is a single, well-scoped operation on an existing/hand-written setup rather than an autoconfig bundle: general guide-driven deploy →
deploy-llm-d;llmdbenchmark-CLI benchmarking (guide or custom workload profiles) →run-llm-d-benchmark; config comparison →compare-llm-d-configurations; WVA/HPA autoscaling →configure-wva-autoscaling-llm-d; teardown →teardown-llm-d. Phases 6 and 7 link to these for the general flows. - Troubleshooting is shared, not duplicated: generic Kubernetes/deploy failures live in
deploy-llm-d's troubleshooting guide;references/pitfalls.mdpoints there and keeps only the autoconfig-specific entries (EPP config, gateway-mode routing, PD/NIXL, modelserver install, benchmark harness).
How this skill is organized
This SKILL.md is the entry point — read it first, then load references/phase-N-*.md for the phase you're working in. Each phase reference is self-contained; you do NOT need to keep prior-phase references in context once a phase is done.
What ships with it
15 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.
- feature_docs.yaml 32 KB
- references/phase-1-cluster-discovery.md 17 KB
- references/phase-2-5-doc-driven-synthesis.md 13 KB
- references/phase-2-discovery-questionnaire.md 33 KB
- references/phase-3-recap.md 5.1 KB
- references/phase-4-call-script.md 15 KB
- references/phase-5-present-recommendation.md 5.9 KB
- references/phase-6-deploy.md 54 KB
- references/phase-7-benchmark.md 14 KB
- references/pitfalls.md 12 KB
- scripts/autoconfig_poc.py 103 KB runs code
- scripts/benchmark.py 28 KB runs code
- scripts/doc_cache.py 14 KB runs code
- scripts/parse_bench_results.py 13 KB runs code
- scripts/verify_doc_map.py 7.7 KB runs code
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.
- 12d ago First seen · 110 lines · 75 tokens per session scan A 2bef7b7a55fc
llm-d-autoconfig is a skill published in the GitHub repository llm-d-incubation/llm-d-skills (6 stars, last pushed 29d ago), licensed Apache-2.0. It adds 75 tokens to every session and 3,548 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
tensorrt-llm
High-throughput LLM inference on NVIDIA GPUs.
google-cloud-solution-guided-gke-ai-migration
Guides the migration of existing AI workloads (Cloud Run, Gemini API, Gemini Enterprise Agent Platform) to self-hosted GKE inference using gcloud and kubectl. Use when the user has an existing AI inference workload (on Cloud Run, the Gemini API, Gemini Enterprise Agent Platform, or a custom VM) and wants to move it to…
agent-platform-tuning
Agent Platform Model Tuning. Use when you need to fine-tune open models or Gemini models using Agent Platform infrastructure. Don't use for model training outside Agent Platform, model deployment to endpoints (use agent-platform-deploy), or managing serving endpoints (use agent-platform-endpoint-management).
gke-inference
Deploys and optimizes AI/ML inference workloads on GKE, using GPUs, TPUs, and model servers. Use when deploying GKE inference servers, configuring GKE GPU resources for inference, or deploying LLMs on GKE. Don't use for generic batch jobs or HPC task queues (use gke-batch-hpc instead).
agent-platform-endpoint-management
Manages Agent Platform serving endpoints. Use when you need to create, list, describe, update, or delete serving endpoints for model deployment on Agent Platform. Also use when troubleshooting endpoint permission, quota, or resource busy errors. Don't use for deploying models to endpoints or for running model…
modal
Modal is a serverless cloud platform for running Python on demand, including on-demand GPUs. Use when deploying or serving AI/ML models, running GPU-accelerated workloads (training, fine-tuning, inference), serving web endpoints, scheduling batch jobs, or scaling Python code to cloud containers with the Modal SDK.