Skills

Skills are reusable instruction packages you write once and apply to any chat message. Each skill has a short name you invoke with a slash command, a description that tells GNG Analyst when the skill applies, and a body of instructions the assistant follows for that turn. Skills carry no files and no code: they are written instructions only. You create a skill, invoke it by typing its slash command, attach it to your subagents, and share it through the Marketplace. Skills are available on every plan, with a per-plan limit on how many you can save.

Before you begin

Sign in to your account. Skills are managed inside the AI Console.
Open Skills from the console sidebar.
Your plan sets how many skills you can save. The Skills surface shows your saved count against your limit.

Quickstart

Open Skills
Open Skills from the console sidebar. The surface opens inside the console workspace, with the sidebar still available.
Create a skill
Select New skill and fill in the skill editor. For every field and its limits, see Create a skill.
Send it to the composer
On the saved skill, select Use in chat. The composer is pre-filled with the skill's slash command, ready for the rest of your message.
Finish and send
Type the rest of your message after the slash command and send. The skill applies to that one message.

What it shows

The Skills surface holds your saved skills and the controls to invoke, edit, and share them. The parts are described in the following sections.

The Skills surface

The heading reads Skills with the caption Your prompt packages. A count line shows how many skills you have saved relative to your plan's limit, or notes that your plan has no limit. When you reach the limit, the line says so and the create action locks. On a wide screen the surface lists each skill in a table; on a phone it becomes a stacked list. Each row shows the skill's name, its slug, its description, an Enabled switch, and the per-skill actions Use in chat, Edit, and Delete. New skill opens the skill editor. Publish to marketplace appears once you own at least one of your own skills. Before you save your first skill, the empty state reads "No skills yet. Create one or import from the Marketplace."

Built-in skills

One built-in skill is provided to everyone: Skill Designer, invoked as /skill-designer, which helps you draft a new skill. Built-in skills are read-only. You cannot edit, disable, or publish them.

The skill picker in the composer

Type / at the start of the message box to open the skill picker, then choose a skill, or type the slug directly, for example /earnings-brief What changed for AAPL this quarter?. Only enabled skills appear in the picker. You can also attach a skill to a subagent from the subagent editor, so the subagent carries those instructions whenever it runs. For that path, see Subagents.

Inputs

Use in chat: pre-fills the composer with the skill's slash command and returns you to the chat, ready to type the rest of the message.
Enabled: a per-skill switch. A disabled skill stays saved but cannot be invoked and is not offered in the picker.
New skill and Edit: open the skill editor. For the fields and their limits, see Create a skill.
Publish to marketplace: lists one of your own skills for other members. The publish flow collects the listing details the same way agent listings do.
Delete: removes a skill permanently.

Outputs and how to read them

A skill shapes how the assistant answers the one message you applied it to: the tone, the structure, and the steps follow the skill's instructions.
Your sent message shows the skill token at the start of the bubble, so the chat records which skill you applied. Anyone viewing a shared copy of the chat also sees the token.
Resending a message that used a skill applies the skill again.

How it is calculated

No formula applies. A skill is written instructions that the assistant reads for one turn. Applying a skill carries no separate charge. The message is metered as usual, by the model, the reasoning depth, and the tools the answer uses. For how metering works, see GNG credits.

Data and timing

Skills are your own saved content, applied at the moment you send the message. Edits to a skill apply to the next message that invokes it. No market data feed is involved.

Use cases

A recurring analysis format: an /earnings-brief skill that structures every earnings question the same way.
A house style: a skill that makes long answers close with a summary table.
A repeatable routine: a checklist the assistant walks through whenever you invoke the skill.
Team sharing: publish a skill so other members can import the same working style.

Limitations and disclosures

The number of skills you can save is capped by your plan, and imported skills count against the same limit. The surface shows your count against your limit.
Skills are instructions only. They hold no files, no code, and no data of their own.
When the assistant loads a skill for you, very long instructions can be trimmed. The editor warns as the instructions grow.
One skill applies per message, at the start of the message. A message routed directly to a single subagent (a pinned one-to-one turn) does not apply skills, and the picker does not open for those turns.

GNG Research provides equity research and educational tools, not investment advice. Nothing on the platform is a recommendation to buy or sell any security. Do your own research and consider your circumstances before making any investment decision.