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/understudylabs/understudy-agent-tools/run-local-model-labnpx skills add understudylabs/understudy-agent-tools --skill run-local-model-labgit clone --depth 1 https://github.com/understudylabs/understudy-agent-toolsWrote 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/understudylabs/understudy-agent-tools/run-local-model-lab)<a href="https://agentmods.dev/skills/understudylabs/understudy-agent-tools/run-local-model-lab"><img src="https://agentmods.dev/badge/skills/understudylabs/understudy-agent-tools/run-local-model-lab.svg" alt="Measured on agentmods" 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 | $0.00082 | $0.02425 |
| Opus 5 | $0.00041 | $0.01213 |
| Sonnet 5 | $0.00016 | $0.00485 |
| Haiku 4.5 | $0.00008 | $0.00243 |
Grade A, and why
run-local-model-lab 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 4d 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 — 168 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Run Local Model Lab
Run a local model against an existing Understudy workload/eval to determine the best route for the objective. Local inference is $0 and private, and it may be the only legal path under ZDR / SOC2 / local-only constraints, but do not use a weak local result to avoid an informative remote anchor when remote evaluation is allowed and could change the decision. Same-family models (e.g. local Gemma 4 → remote Gemma 4 31B via the gateway) graduate cleanly.
Apple Silicon + MLX only. On Macs, MLX is the native local path — quantized
open weights against unified memory at the best tokens/sec, no GPU drivers, no
build step. This skill standardizes on mlx_lm.server (an OpenAI-compatible
endpoint, one model per port); it does not use Ollama or llama.cpp.
Want to meet the model, not just score it? Use
../ladder/SKILL.md for the no-data onboarding climb:
it opens the local gemma-4-e2b lane in a browser, streams scored tasks, and can
optionally compare against the billed gateway lane. Keep this skill for measured
runs against the user's real workload.
This skill measures and recommends; it does not download weights or change
production routing on its own. To compare several candidate models (any mix of
local, gateway, frontier) on one frozen eval, use
../compare-model-sweep/SKILL.md.
When to use
A workload already has (or can get) a frozen eval — see
../capture-evidence/SKILL.md — and the developer
wants a local candidate evaluated before remote spend, or needs a local-only
route for compliance. For pure remote inference/routing use
../use-understudy-gateway/SKILL.md.
Safety Gates
- No weight downloads without explicit approval and a stated size cap. Model weights are large; confirm the exact model + quantization + disk size first.
- Local-first, no upload. Keep traces, prompts, and outputs local unless the developer approves a specific upload. This is the compliant path — do not break it.
- Gated weights (Gemma, etc.) need license acceptance + an HF token; never print or commit the token.
- Never evaluate an
-assistantdrafter on its own. The*-it-assistantmodels are speculative-decoding drafters (MTP), not standalone models — they only speed up a paired target while preserving its quality. Seereference.md.
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.
- 4d ago First seen · 168 lines · 82 tokens per session scan A 82adde94e293
run-local-model-lab is a skill published in the GitHub repository understudylabs/understudy-agent-tools (16 stars, last pushed 2d ago), licensed MIT. It adds 82 tokens to every session and 2,425 once invoked, about $0.0004 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
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
auto-perf-optimize
Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.
chat-perf
Run chat perf benchmarks and memory leak checks against the local dev build or any published VS Code version. Use when investigating chat rendering regressions, validating perf-sensitive changes to chat UI, or checking for memory leaks in the chat response pipeline.
chat-pet-sprite-creation
Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.
cpu-profile-analysis
Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…