Borrowing it
Nothing to install: this file belongs to cloudrift-ai/emmy. 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/cloudrift-ai/emmy/main/.agents/skills/discover-models/SKILL.mdgit clone --depth 1 https://github.com/cloudrift-ai/emmyWrote 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/cloudrift-ai/emmy/discover-models)<a href="https://agentmods.dev/skills/cloudrift-ai/emmy/discover-models"><img src="https://agentmods.dev/badge/skills/cloudrift-ai/emmy/discover-models/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/cloudrift-ai/emmy/discover-models"><img src="https://agentmods.dev/badge/skills/cloudrift-ai/emmy/discover-models.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00077 | $0.02079 |
| Opus 5.5 | $0.00031 | $0.00832 |
| Sonnet 5.5 | $0.00015 | $0.00416 |
| Haiku 4.5 | $0.00008 | $0.00208 |
Grade A, and why
discover-models 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 16d 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 — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Discover Models to Explore
Turn "what models are worth our GPU hours?" into a shortlist or a refresh of Emmy's recipe lifecycle. New open-weight models are filtered to the ones with real demand and mapped to the GPU platforms they fit. Every recipe gets a 0-100 heat score for current onboarding priority; the maintained set stays small and focused, other useful recipes stay best-effort, and only technically superseded or unusable models become obsolete.
Research is keyless and read-only: public OpenRouter and Hugging Face endpoints plus web search. No server is touched
and no recipe is edited — the workflow applies a validated lifecycle manifest, and onboard-model owns deployment
and qualification. In automated mode, the agent may create or edit only recipe-local DISCOVERY.md notes to retain
research behind its decisions.
HF/OpenRouter data ─┐
Reddit discussions ─┼→ reconcile exact models → score heat → VRAM fit → GPU platform → hardware/model matrix
OpenRouter/Arena ───┘
Choose the mode
Automated lifecycle mode (GitHub Actions): read prompts/discover-models/lifecycle.md and
prompts/discover-models/score-recipes.md and the recipe-local DISCOVERY.md notes supplied in the task before
research. lifecycle.md is the whole contract — task payload, delegation to the source, scorer, and fit subagents,
selection rules, and output JSON — and this skill adds only the background below. Ask no questions, never rebuild the
inventory the task supplies, and return the selection as soon as the evidence supports it. The workflow runs the agent
on a checkout of main and commits the validated manifest and any changed research note itself; the agent never
touches git.
Survey mode (interactive): produce a shortlist and a hardware/model matrix. Ask only what the user has not implied:
- Time window — default the script's ~90 days; widen with
--since 2026-01-01. - Modality — include multimodal by default;
--text-onlyif they serve only text. - Target hardware — default every GPU in
emmy/gpu.py; bucket only for a subset they name. - Finalists — default 5-8, spread across hardware tiers.
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.
- 16d ago Changed · +3 lines 3cd7f277fe90
- 18d ago Changed · +2 lines d26aefb6a1e1
- 20d ago Changed 45a9229858af
- 24d ago Changed · -100 lines ccc0a91ce9a8
- 1mo ago Changed 28c3aa039afb
- 1mo ago First seen · 244 lines · 77 tokens per session scan A 5678d44a6756
discover-models is a skill published in the GitHub repository cloudrift-ai/emmy (116 stars, last pushed today), licensed Apache-2.0. It adds 77 tokens to every session and 2,079 once invoked, about $0.0003 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 skills, from other repositories
inference-aiops
Use this skill whenever the user needs to operate a GPU inference cluster — vLLM (OpenAI API + Prometheus /metrics) and Ray Serve / Ray Jobs (Ray dashboard), plus the single-process serving engines SGLang and TGI (Text Generation Inference): a one-shot cluster overview (deployments + total replicas + queue…
sglang-diffusion-performance
Use when choosing the fastest SGLang Diffusion flags for a model, GPU, and VRAM budget.
sglang-diffusion-add-model
Use when adding a new diffusion model or Diffusers pipeline to SGLang.
sglang-diffusion-cache-dit
Workflow for upgrading/integrating cache-dit in SGLang diffusion (multimodalgen): DBCache, DMD calibrator, TaylorSeer, SVDQuant DQ; porting upstream PRs and resolving conflicts against the per-request knob system; adding new cache knobs; building the sglang generate CLI test matrix; precision validation (PSNR / log…
sglang-diffusion-modelopt-quant
Use when quantizing a diffusion DiT with NVIDIA ModelOpt and making the resulting FP8 or NVFP4 checkpoint loadable, verifiable, and benchmarkable in SGLang Diffusion.
hqq-quantization
Half-Quadratic Quantization for LLMs without calibration data. Use when quantizing models to 4/3/2-bit precision without needing calibration datasets, for fast quantization workflows, or when deploying with vLLM or HuggingFace Transformers.