Dynamic skill scanning
Automatically loads skills from project .claude/skills/ and global ~/.claude/skills/ directories, extracting titles, descriptions, and icons — no code changes required.
Raycast extension for Claude Code Skills
Execute Claude Code Skills from Raycast
Dynamic skill scanning, one-click execution, real-time status tracking, and structured logging — AI Agent power in your launcher.
Why Agent Executor
Agent Executor is a Raycast extension that brings the Claude Code Skills ecosystem into your launcher. It automatically scans .claude/skills/ directories, dynamically loads all skills, and lets you execute them with one click while tracking status in real time. No hardcoding needed — add a new skill by simply creating a subdirectory with a SKILL.md file.
Core capabilities
Automatically loads skills from project .claude/skills/ and global ~/.claude/skills/ directories, extracting titles, descriptions, and icons — no code changes required.
Select a skill, add an optional instruction, and hit Enter. Supports both background and terminal-popout modes, with base64-encoded prompts to prevent injection.
Auto-grouped by Running / Completed / Failed. Running tasks refresh every 5 seconds — quickly spot what needs attention.
JSONL-format execution history (Run ID, PID, elapsed time, full output, exit code) retained for 7 days, viewable directly inside Raycast.
Concurrency control with configurable max parallel tasks (1–10). Queue survives window closure.
Configure up to 5 project directories and manage skill sets across multiple Agent workspaces.
Workflow
Install
This extension requires Raycast and Claude Code CLI. Clone the repo, run npm install && npm run dev, and it appears in Raycast. See the README for detailed setup instructions.