Choose model lab workflow

Skill gaelic-ghost/socket/plugins/model-lab-skills/skills/choose-model-lab-workflow

Route language-model training, data, evaluation, checkpoint, representation, steering, ablation, jailbreak, tool-calling, Apple-runtime, and benchmark requests. Use when the primary workflow or Socket owner is unclear.From its SKILL.md

Install

npx -y skills add gaelic-ghost/socket --skill choose-model-lab-workflow

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.1 KB, 616 tokens by cl100k_base, as published. Nobody here has run it

Choose Model Lab Workflow

Outcome

Select one primary workflow, name any supporting workflows, and make the evidence boundary explicit before work begins.

Route The Request

Requested outcome	Primary skill
Define a hypothesis, controls, budget, and artifacts	`design-model-experiment`
Curate, transform, split, or document examples	`prepare-language-model-dataset`
Run SFT, LoRA, QLoRA, or a full parameter update	`fine-tune-language-model`
Measure capability, behavior, quality, or safety	`evaluate-language-model`
Decide which checkpoint is better and why	`compare-model-checkpoints`
Choose Core AI, Core ML, MLX, ExecuTorch, or Foundation Models	`choose-apple-model-runtime`
Locate or test internal representations	`research-model-representations`
Apply activation or weight-space behavior steering	`steer-language-model-behavior`
Remove or suppress a refusal direction	`ablate-refusal-representations`
Measure jailbreak or prompt-injection robustness	`evaluate-jailbreak-resilience`
Measure tool selection, arguments, execution, or recovery	`evaluate-tool-calling-model`
Compare latency, memory, energy, throughput, or artifact size	`benchmark-model-runtime`
Run preference optimization such as DPO/ORPO	Keep the experiment and eval here; use the supported TRL workflow through `fine-tune-language-model` until a stable standalone skill is earned
Pretrain or continue pretraining a foundation model	Do not collapse it into fine-tuning; define the distributed/corpus contract and treat `train-language-model` as a deferred skill candidate
Merge adapters or quantize/package an artifact	Use `compare-model-checkpoints` around the exact transformation; use the project-native tool and evaluate the deployable output
Evaluate an agent skill, plugin, or host harness rather than a model protocol	Hand off to `productivity-skills` and `agent-portability-skills`

Respect Ownership Boundaries

Use cloud-inference-skills for provider, GPU, endpoint, cost, and teardown decisions.
Use python-skills for Python packaging and environment repair.
Use apple-dev-skills for Swift/Xcode application integration after the runtime has been chosen.
Use productivity-skills for evaluating agent skills, prompts, or plugin packages rather than model checkpoints.
Use cybersecurity-skills when an authorized evaluation targets a deployed system instead of a model artifact.

Return A Routing Contract

State:

the primary skill;
supporting skills in execution order;
the controlled variable;
the artifact or metric that proves completion;
any paid compute, data-access, or deployment authorization required.

Do not silently turn research planning into a paid run, model publication, or production-system test.

What ships with it: 1 file

238 B alongside SKILL.md

agents/

openai.yaml238 B

Gives 0 of the 12 instructions most automation workflows skills give in 616 tokens

Counted across 745 of the 1,008 authors here whose files we hold, read 2026-08-07

Write conventional commit messagesin 36 of 745, across 35 files
Delete branches after mergein 30 of 745, across 21 files
Make atomic commitsin 25 of 745, across 15 files
Write minimal code to pass testsin 22 of 745, across 10 files
Re-snapshot after navigation or DOM changesin 21 of 745, across 13 files
Use try-catch for error handlingin 20 of 745, across 8 files
Run tests before committingin 20 of 745, across 12 files
Write tests before implementationin 20 of 745, across 8 files
Configure branch protection rulesin 19 of 745, across 5 files
Explain the why in commit messagesin 19 of 745, across 9 files
Refactor code while tests remain greenin 19 of 745, across 6 files
Interact with elements using refsin 19 of 745, across 11 files

Said here and by no other author read

select one primary workflow
name supporting workflows in execution order
state the controlled variable
name the artifact or metric proving completion
list required compute, data, or deployment authorizations
hand agent evaluation requests to productivity skills

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.