Agent Skills for Gemini CLI

Gemini CLI is good at holding a lot of context. Skills keep that context clean: instead of pasting a 900-word brief every time, you install a folder that activates only when the task matches. The agent reads the name and description up front, then loads the body on demand.

Where Gemini CLI looks for skills

ScopePathNotes
Workspace.gemini/skills/<name>/SKILL.mdShared with the repo
Workspace (alias).agents/skills/<name>/SKILL.mdCross-agent standard
User~/.gemini/skills/<name>/SKILL.mdAvailable everywhere
User (alias)~/.agents/skills/<name>/SKILL.mdWins over .gemini at the same tier

Within the same tier, the .agents/skills alias takes precedence. Don't keep two copies of the same skill name — you'll spend an afternoon debugging which one loaded.

Install in 30 seconds

  1. Copy the skill folder to .agents/skills/.
  2. Check the path ends with <name>/SKILL.md.
  3. Start a new session and ask for the output you want.
  4. On first use, Gemini CLI asks for consent before loading the skill's files — read the path it shows.

Skills that suit long-context work

SkillWhat it producesInstalls
Chalkboard InfographicA single-page explainer graphic267
Image Narrative DeckA story-led slide deck171
Article Brief DeckA briefing deck from a long article154
Research Talk DeckAn academic talk deck134
Summary Line ChartA clean trend chart53
Article Illustration SetThree consistent illustrations58

A workflow that fits the CLI

Long documents are the point. Feed a PDF or a repo, ask for a briefing deck, then ask for the OG image and the summary chart in the same session. Three skills, one context, no re-explaining.

FAQ

Do skills count against the context window? Only the metadata is always present; the instructions load when activated.

Can I use Claude Code skills here? If they're plain SKILL.md folders, yes — the format is shared.

Where do I start? Article to deck or OG image.