Source: Agent Skills (Gemini CLI docs)

Gemini CLI's implementation of Agent Skills: four discovery tiers, model-called activate_skill with a user consent prompt, and /skills and gemini skills management commands.

Contents
  • What: official product documentation
  • Publisher: Google (Gemini CLI project)
  • Date: living document, read 2026-11-14
  • Original: Agent Skills

Summary

Shows a different activation mechanism from Claude Code: the model calls a dedicated activate_skill tool and the user approves before the body enters context.

Key claims

  • Tiers (low to high precedence): built-in, extension, user (~/.gemini/skills/ or ~/.agents/skills/), workspace (.gemini/skills/ or .agents/skills/); .agents/ wins over .gemini/ within a tier.
  • Startup: name and description of each enabled skill go into the system prompt.
  • Activation: the model calls activate_skill; a confirmation shows the skill’s name, purpose and directory; on approval the body and folder structure are injected and the directory becomes an allowed path.
  • Management: /skills list|link|enable|disable|reload; gemini skills install <git url|path> with --scope and --consent (skips the security confirmation).

Feeds

Agent skills

Suggest changes →