/skills opens a browser of every skill available to the current session. Selecting one
inserts it into the prompt box as /<skill-name> so you can add arguments before sending.
A skill is a packaged set of instructions for a particular kind of task — a review checklist, a report format, a domain workflow. Loading one gives the agent that procedure in full, instead of relying on it to improvise.
The agent can load skills itself. Every skill's description is presented to the model, and
it calls the skill tool when a task matches one. Reach for /skills when you want to:
If the agent should have picked a skill and didn't, the usual cause is its description — see below.
Skills are discovered from two directories, and both are merged into one list:
| Location | Scope |
|---|---|
~/.latentcode/skills/ |
your personal skills |
~/.config/latentcode/skills/ |
config-directory skills |
Additional folders can be added with the skills key in latentcode.json.
Tip
Having skills in both directories works but makes them hard to keep track of. Consolidating on one location is worth doing early.
A skill with no description in its front-matter is invisible to the agent. The model
chooses skills by reading descriptions; an empty one cannot be matched, so the skill will only
ever run if you invoke it by hand.
---
name: debt-serviceability-analysis
description: Assess whether a borrower can service and repay a debt facility — DSCR,
redemption schedule, cash available for debt service. Use when asked about debt
serviceability, repayment capacity, or "can this loan be repaid".
---
Write the description as when to use this, not what this is. Trigger phrasing — the words a user would actually type — matters more than a tidy summary.
/ menuIndividual skills are deliberately excluded from / autocomplete. With a large skill library
they would swamp the built-in commands and make the menu unusable. /skills is the way in.
Typing /<skill-name> directly still works if you know the name.