Triage Ml Task by probabl-ai — AI Skill for Any LLM

Overview

Publisher	probabl-ai
Repository	`skore-skills`
Skill name	`triage-ml-task`
Stars	144
Forks	9
Bundled files	1
License	BSD-3-Clause
Links	Skill source Repository Homepage

Markdown instructions

A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

Works with any LLM

AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

1 bundled files

Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

Open source

Published by probabl-ai on GitHub. Read the source before you install it.

Installation

Install the Triage Ml Task AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

Open the app and go to Plugins → Skills.
Choose "Install from GitHub".
Paste the skill folder URL below and confirm.
https://github.com/probabl-ai/skore-skills/tree/main/skills/triage-ml-task
Enable the skill in any chat where you want it available.

Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills

git clone --depth 1 https://github.com/probabl-ai/skore-skills.git /tmp/skore-skills
mkdir -p .claude/skills
cp -r /tmp/skore-skills/skills/triage-ml-task .claude/skills/triage-ml-task

Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Triage Ml Task in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Triage Ml Task on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Triage Ml Task is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

Loaded only when it is needed

The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

Switch models mid-chat

Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Triage ML Task

This is the session owner. Stage skills do the work; you only route and ask. Do not execute another skill's methodology.

Human-facing prose

Details: setup-workspace references/human_facing_prose.md. Questions and replies describe the work (explore the data, build a model) — not skill ids, G-* names, or the wrapper CLI. Run python -m skore_skills … yourself; do not quote it.

Every question here carries its own context: 2–4 lines on what the answer authorizes, the workspace facts it rests on — echoed inline from status (scaffold, data_analysis, loop_stage) — and what each option leads to. A file link is an addition, never the context.

Procedure

Run python -m skore_skills status. Read skills, data_analysis, loop_stage, setup.pending, and the filesystem snapshot. If .skore is missing, that is expected. Do not treat a missing file as an empty project when src/ or journal/ exist.

Certain request — load that skill, after the setup gate below. Tell the user the work you are starting, not the catalog id. Do not list the catalog. status.skills is a per-id dict. Load the mapped skill only if status.skills.<id> is true; else one-line skip and do not invent that skill's steps:

Setup gate for a lifecycle skill. Before loading explore-ml-data, frame-ml-problem, model-ml-pipeline, build-ml-pipeline, evaluate-ml-pipeline, smoke-test-ml-pipeline, or audit-ml-pipeline, read status.setup.pending. If it is non-empty and status.skills.setup-ml-project is true, load setup-ml-project and stop. Tell the user the requested work waits on those pieces. Do not show the entry menu. After setup returns, load the certain lifecycle skill. Do not load setup again on this turn. If setup.env or setup.workspace is declined, stop in one line. A declined git or editable does not block. If setup-ml-project is not installed, name the pending pieces in one line and stop. Do not invent git init, scaffold, or env init. Uncertain sessions skip this gate.

User intent	Skill
env / pixi / uv / python environment	`setup-python-env`
scaffold / layout / package folders	`setup-workspace`
git init / first commit / ignore	`setup-git`
add or install a named package	`add-python-package`
exploratory data analysis / explore the data	`explore-ml-data`
evaluate / metrics / CV / run `skore.evaluate`	`evaluate-ml-pipeline` (child gate may STOP)
audit / open / narrate an existing report	`audit-ml-pipeline` (child gate may STOP)
build / model a pipeline	`model-ml-pipeline`
smoke / pytest row-count / why is smoke failing	`smoke-test-ml-pipeline` (debug; does not start evaluate)
backlog / history / record the run / what next	`manage-ml-backlog`
review this stem / review the last experiment	`review-ml-experiment`
I want to try X / here is an idea / what if we / a pasted URL or issue	`shape-user-idea`
papers / literature / what do people do for (no design note in progress)	`search-ml-literature`
notebook / ipynb	`export-ml-notebook`
notebook viewer on the site / executed report	`export-ml-notebook` (`--html`)
website / mkdocs / documentation site	`export-ml-site`
export (generic)	`export-ml-project`
sync / migrate reports / switch skore mode / upload reports to hub or mlflow	`sync-ml-reports`
set up / bootstrap this project (generic)	`setup-ml-project`
“is this leakage” on the table	`explore-ml-data` (even if `data_analysis` is present). Do not load `research-ml-practice`.
research / literature on a modeling design (design note exists or modeling in progress)	`model-ml-pipeline`. Do not load `research-ml-practice`.
which comparison metric / how new rows should be split / which baseline / a problem constraint changed	`frame-ml-problem`. Not a request to run evaluation.
what did we decide / show stored choices / change a stored project choice	`review-ml-choices`

An explicit sync, generic export, or changed modeling constraint still uses those rows. Do not send them through review-ml-choices.

A git init, first commit, or ignore request loads setup-git and stops. Do not write git init, a .gitignore, or a commit plan. With no shell, the whole answer is that setup-git is loaded. Do not describe the commands that skill will run.

"What should we try next?" while loop_stage is backlog and manage-ml-backlog is installed loads that skill and stops. Do not ask the user to choose explore, build, review, or export.

Certain EDA: run python -m skore_skills status. If status.setup.pending is non-empty, load setup-ml-project (setup gate above) and stop. Otherwise load explore-ml-data and stop. Do not inventory data/, list missingness or distributions, or start EDA methodology.

Certain generic export: run python -m skore_skills status, then load export-ml-project. Do not list notebook / --html / site as sibling options.

Modeling while status.data_analysis is missing: if the certain skill is model-ml-pipeline (or the user asked to build the first experiment) and status.skills.explore-ml-data is true, do not load modeling yet. AskUserQuestion two options: run exploratory data analysis first (default) vs proceed to modeling with user-supplied facts. Do not invent dataset facts here. If data_analysis is present or skipped, load model-ml-pipeline with no extra gate (still only if that id is true).

Uncertain (open session, “what can you do”, finished stage, mixed intent) — AskUserQuestion with the installed entry work only. One pick, then load the mapped skill.

Offer a label only if status.skills.<id> is true. User-visible labels (no ids):
- Set up the project → setup-ml-project
- Explore the data → explore-ml-data
- Build a model → model-ml-pipeline
- Review the last experiment → review-ml-experiment
- Record / decide what next → manage-ml-backlog
- Export → export-ml-project
- Sync reports → sync-ml-reports
- Review choices → review-ml-choices
If none of those ids are true, say so in one line; do not invent a menu.

Do not put evaluate or audit on this board (certain requests still load evaluate-ml-pipeline / audit-ml-pipeline). Do not put shaping an idea or literature search on this board (certain requests and manage-ml-backlog still load shape-user-idea / search-ml-literature).

If status.data_analysis is missing and status.skills.explore-ml-data is true, recommend exploring the data first. Do not auto-load it. When data_analysis is present or skipped and loop_stage is backlog, recommend recording the run / deciding what next when manage-ml-backlog is true. Do not auto-load it.

Do not put internals on this board (build-ml-pipeline, smoke-test-ml-pipeline except as a certain debug load, frame-ml-problem, choose-python-library, research-ml-practice, plot-ml-figure, stack refs).

Stop conditions

Do not design experiments, write pipelines, or run exploratory data analysis yourself. Certain EDA with setup.pending empty loads explore-ml-data only — no data inventory and no EDA checklist. Pending setup loads setup-ml-project first.
Do not load every skill.
Do not invent workspace facts when status is unavailable.
Do not treat a missing .skore as an empty project when src/ or journal/ exist.
Do not invent a missing skill's steps from memory.
Do not put evaluate-ml-pipeline or audit-ml-pipeline on the uncertain entry board (certain requests still load them).
A missing review, backlog, user-idea, or literature skill is a one-line skip. Do not invent that skill's procedure.
Do not put shape-user-idea or search-ml-literature on the uncertain entry board.
Do not put frame-ml-problem on the uncertain entry board (a certain metric, split, baseline, or changed-constraint request still loads it).
Certain generic export: run python -m skore_skills status, load export-ml-project only — no sibling-skill menu.

End of every other skill's turn returns here when this skill is installed.

Frequently asked questions

What does the Triage Ml Task AI skill do?

Session owner: list installed entry skills and ask which to run. Load a skill without asking only when the request is certain to be that skill. Trigger on an ambiguous request, a finished stage, a workspace-open session, or "what should we do next".

Why use Triage Ml Task on TypingMind?

Because you install it once and use it with any model. Triage Ml Task is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Triage Ml Task in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/probabl-ai/skore-skills/tree/main/skills/triage-ml-task. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Triage Ml Task?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Triage Ml Task?

As many as you like. As long as a model supports skills, you can use Triage Ml Task with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Triage Ml Task AI skill free?

Yes. It is published on GitHub by probabl-ai under the BSD-3-Clause license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

Overview

Installation

Install in TypingMind

Install in another agent

Use it in TypingMind

Works with any AI model

Skill instructions

Triage ML Task

Human-facing prose

Procedure

Stop conditions

Bundled files

Frequently asked questions

Related AI Skills

Set up your own AI workspace now