sg-model

Choose the right model for the task without letting speed or cost outrank quality.

A model-selection skill for matching task type, reliability, quality bar, latency, cost, and subagent routing needs before execution starts.

Browse skills on GitHub Back to the hub

Problem

What problem this solves

Using the wrong model can waste money, time, or execution reliability, but choosing the cheapest or fastest model too early can also weaken important work.

Outcome

What changes when it works

You get a model choice that starts from reliability and risk, then uses speed or cost only when the fallback remains quality-equivalent.

Why It Matters

What this changes for a solo founder

This skill matters when model choice is no longer trivial. It keeps the workflow economical without treating economy as more important than correctness, security, maintainability, and proof.

When to use it

When the task could reasonably fit several model tiers
When cost or latency matters materially but should not lower the quality bar
When a complex task may justify stronger reasoning

What you give it

A task description and its constraints
Any quality, reliability, cost, or speed constraint

What you get back

A model recommendation
A clearer tradeoff between reliability, speed, cost, and reasoning depth
Quality-equivalent fallbacks instead of automatic cheapest-path defaults
A default subagent model policy: low-risk bounded work can use GPT-5.4-mini, long implementation uses GPT-5.3-Codex, high-risk reasoning uses GPT-5.5
A clear note on whether the model choice is applied, a subagent override, or only a next-run recommendation
A better execution setup before work begins

Typical examples

/sg-model for migration planning
/sg-model docs cleanup without lowering quality
/sg-model which model for this audit

Limits

It helps choose a model; it does not improve the task framing by itself
It cannot guarantee that the already-running main conversation can switch its own active model mid-thread
The best model still depends on the quality of the downstream workflow
Fast or cheap fallbacks are only appropriate when they remain quality-equivalent for the task risk

Related Skills

See the rest of the loop around this move.

Plan & Decide

sg-context

Prime the relevant files and decisions before the thread starts guessing about the codebase.

A context-loading skill that retrieves the most relevant repo surface before implementation begins.

Open skill page

Plan & Decide

sg-spec

Turn a fuzzy request into an implementation-ready contract before code starts mutating the repo.

The spec-writing skill for shaping non-trivial work into a clear user story, behavior contract, executable task list, and chantier history.

Open skill page

Build & Fix

sg-start

Start implementation when the task is ready, and force a stronger path when it is not.

The main execution entrypoint for beginning real work after the problem and contract are clear enough, while keeping the chantier flow visible.

Open skill page

Go Deeper

This public page explains the move. The repository explains the execution contract.

Use the public catalog for education and workflow selection. Move to the docs and GitHub when you want the operator-grade execution details behind the skill system.

Documentation Overview Open skills on GitHub Return to the skills hub