Skills

Skills: find and use reusable instructions

Find library and community skills, pick instructions for a message, pin them in a workspace, and understand automatic selection and answer badges.

Before you start

A skill is reusable instructions for a kind of task, stored in SKILL.md. It guides an answer’s process, tone or format; it is not a model, a saved conversation, or permission to use a connected app. Templates create a workflow of nodes; skills guide how an answer is written.

Open Skills from your avatar menu when it is available for your account. You can also open Manage skills… from the Skills button in a message box. Availability depends on account access and enabled features. The public gallery is at the app’s /skills page when the gallery is enabled; reading its public pages does not require signing in. Adding a community copy to your account does.

When to use this

Use a skill for repeatable work such as meeting notes, email replies or code review. Pick it yourself for one branch, or pin it when its instructions should guide ordinary answers throughout a workspace. Use automatic selection only when you want Jev to decide whether an eligible skill fits each message.

Read unfamiliar instructions before using them. Community content and shared files can contain misleading instructions; publication review is not a guarantee of safety or accuracy.

Steps

Discover a skill. On the public Skills gallery, search or choose All, By Neurothread, or Community. Open a card and read Instructions, When it is used, and any Files of the skill. Download gives you the Markdown or, for a skill with resources, a ZIP.
Bring it into the app. Choose Use in Neurothread, sign in if needed, and review the preview. For a community skill, choose Add to my skills; its copy starts as Only when I pick it. For a library skill, Open in Explore opens the existing library entry—there is no installation or new personal copy to make unless you want to customize it.
Browse the built-in library. In the signed-in Skills window, Explore contains Neurothread’s library, not the whole community gallery. My skills contains your own creations and copies. Search by title, name or description. Library skills are already available; use Duplicate and edit only if you need your own editable version.
Choose its use mode. On a skill card, the Use selector offers Automatic, Only when I pick it, and Off. Automatic entries are candidates, not instructions applied to every message. Off entries disappear from the manual picker and automatic candidate list. Turning a card off does not clear an existing workspace pin or inherited manual selection; an already-selected skill can still be applied. Remove its chip and unpin it separately.
Pick instructions manually. In Chat’s composer or an ordinary Canvas draft, open Skills, search, and tick the skills you want. Up to three can be selected. Chosen skills appear as chips; use a chip’s × or untick the entry to remove a manual selection. Selections carry into the next messages of that branch until you change them. Send or generate the message as usual.
Pin to the active workspace. Open the workspace, then Skills → Manage skills… or avatar → Skills. On a card, choose Pin in this workspace. The button becomes Pinned in this workspace; click it again to unpin. This control is in the Skills window, not a workspace-settings page. The message picker shows a pin icon beside pinned entries but does not create or remove pins itself. Without an active workspace, the card has no pin button.
Use automatic selection if offered. Fresh Canvas/Chat conversations start with automatic skills off; ordinary child branches inherit their parent’s switch and manual choices. When it is available for your account, tick Automatic skills in Chat or Auto in a compact Canvas draft. Set at least one skill to Automatic first. A Jev adds: … suggestion may appear while typing; click it to make that skill a manual choice, or untick the switch to avoid an automatic addition. The switch, suggestions and Test activation are hidden when the skills decision feature is disabled.
Check the answer. Its skill badges show which instructions actually entered that answer. Hover for Picked by you, Pinned in this workspace, or Chosen by Jev with a percentage. Click a badge to open the skill. Removing a chip affects future messages, not an answer already generated; to stop a pinned instruction, unpin it separately.

Worked example

Open Explore and find the meeting-notes skill. Keep Automatic skills off, select that skill with the composer’s Skills button, and ask: “Turn these notes into decisions, owners and next steps.” Check the answer’s badge for Picked by you.

For a workspace dedicated to meeting records, pin the skill from its card. A later message in that workspace can use it even with no manual chip and automatic skills off. For an unrelated task, unpin it first rather than only removing the chip. No extra library installation is required.

What happens next

For ordinary Canvas/Chat answers, selection priority is manual → workspace pins → automatic. Manual choices suppress Jev’s automatic choice for that message; pins can still be added beside manual choices. Without a manual choice, Jev may add one matching automatic candidate, or none. Pins work independently of the automatic checkbox. Skills are deduplicated and only those that fit the answer’s limits are included.

Continue on a cut-off answer starts a new answer with all that answer’s saved skills as manual choices and automatic selection off. This can reuse a skill even after its card was turned Off or its workspace pin removed. It differs from ordinary branching; check the reused choices before sending later messages.

Agents use a separate mechanism. In Agent settings → Tools, Use skills (your automatic skills, loaded when a task matches one) is enabled by default when skills are available. With a tool-capable model and eligible entries, the agent sees an index of automatic skills and can load instructions through use_skill. This is not the Canvas/Chat Jev checkbox and does not require that checkbox to be on. Workspace pins are not the agent’s skill index. Skills do not enable the computer, connected apps, or their permissions.

Cost and limitations

Applied skill text adds input tokens to the chosen model’s context; the answer follows normal credits or provider billing. Longer skills and multiple pins can increase usage. The editor’s token count is an estimate, not the final bill. Jev’s helper selection is not separately charged as an answer in the current implementation; that is not a promise of permanent free service.

An ordinary answer includes at most three skills, subject to model context capacity and a shared size budget for pinned/automatic instructions. Manual choices take priority but still need to fit the model. Avoid contradictory instructions. Skills are not applied to Build generation, decision nodes or critique answers; an ordinary question about a Build project is a different case. Canvas/Chat uses the skill’s instruction text, not automatic execution of resource files. Agent indices are capped and resource reads can be truncated.

Your personal skill allowance is shown in My skills. The library does not count toward it; personal copies do. Availability and limits can change, so use your account’s displayed allowance rather than a fixed plan quota from a guide.

Troubleshooting

Skills is missing: access may be disabled for your account, or its access check failed. Reopen the app and check your connection. A missing automatic checkbox alone does not mean manual skills are unavailable.
A skill is missing from Explore: your own skill with the same name replaces the library entry for you. Look in My skills. Explore does not include community publications.
A skill is absent from the picker: clear the search and check its Use mode. Change Off to Only when I pick it or Automatic. Deleted selections are cleared after the list reloads.
Removing a chip did not stop the skill: check for a workspace pin or an enabled automatic switch. Unpin separately; Off alone does not remove a pin.
No pin button appears: open a workspace in the app before opening Skills. The public gallery and agent page are not workspace pin controls.
Jev adds nothing: the feature may be disabled, no eligible skill may fit, confidence may be low, the message may be too short, or selection may have failed. Pick a skill manually if you need it.
A pin has no answer badge: the skill may have been dropped because other selections filled the three-skill limit or its text did not fit. Shorten instructions or reduce selections.
An agent does not load a skill: check its separate Use skills switch, the skill’s Automatic mode, account access, and model tool support. Do not look for the Chat checkbox in agent settings.

For agents, continue with these Spools guides: