/zed-task

allowed-tools: Read, Write, Bash(test:*), Bash(eza:*), Bash(fd:*)
name: "Zed Task"
description: "Generate optimized Zed task configuration with intelligent defaults"
author: "wcygan"
tags: ["tool"]
version: "1.0.0"
created_at: "2025-07-14T00:00:00Z"
updated_at: "2025-07-14T00:00:00Z"

Context

• Session ID: !`gdate +%s%N`
• Task description: $ARGUMENTS
• Current directory structure: !`eza -la --tree --level=2 | head -20`
• Existing .zed directory: !`test -d .zed && echo "exists" || echo "not found"`
• Current tasks file: !`test -f .zed/tasks.json && echo "found" || echo "not found"`
• Project type detection: !`fd -t f "(deno.json|package.json|Cargo.toml|go.mod|pom.xml)" -d 1 | head -5`
• State file: /tmp/zed-task-state-!`gdate +%s%N`.json

Your task

STEP 1: Initialize session state

• Load or create session state file
• Parse task description from $ARGUMENTS
• Determine project type and appropriate defaults

STEP 2: Analyze existing Zed configuration

• Check if .zed directory exists
• IF .zed/tasks.json exists:
• Read current configuration
• Identify existing task patterns
• Avoid duplicate task labels
• ELSE:
• Prepare to create new tasks.json

STEP 3: Generate intelligent task configuration

• Parse task description to extract:
• Task label (descriptive name)
• Command to execute
• Working directory (default: $ZED_WORKTREE_ROOT)
• Terminal requirements
• Concurrency settings
• Reveal behavior

• FOR EACH common task type:
• Test runners: Use file-specific patterns ($ZED_FILE)
• Build commands: Project root with appropriate builders
• Linters/formatters: File-specific with watch mode
• Development servers: New terminal with auto-reveal
• Script runners: Inherit project environment

STEP 4: Optimize task configuration

• Use modern command alternatives:
• ripgrep (rg) instead of grep
• fd instead of find
• eza instead of ls
• bat instead of cat

• Apply Zed variable best practices:
• $ZED_WORKTREE_ROOT for project operations
• $ZED_FILE for file-specific tasks
• $ZED_COLUMN/$ZED_ROW for cursor-aware operations

STEP 5: Create or update tasks.json

• IF .zed/tasks.json doesn't exist:
• Create .zed directory
• Generate new tasks.json with task array
• ELSE:
• Merge new task into existing configuration
• Preserve existing tasks
• Maintain JSON formatting

STEP 6: Provide usage instructions

• Show task configuration in formatted JSON
• Explain Zed keyboard shortcuts:
• `cmd-shift-p` → `task: spawn`
• `task: rerun` for last task
• Suggest keybinding creation for frequent tasks
• Document Zed variable usage

STEP 7: Save session state

• Record generated task configuration
• Update state file with completion status
• Clean up temporary files

Example Output Format

{
  "label": "Run tests",
  "command": "deno test",
  "cwd": "$ZED_WORKTREE_ROOT",
  "use_new_terminal": false,
  "allow_concurrent_runs": false,
  "reveal": "always",
  "env": {
    "DENO_NO_PROMPT": "1"
  }
}

Task Type Templates

**Test Runner:**

{
  "label": "Test current file",
  "command": "deno test $ZED_FILE",
  "cwd": "$ZED_WORKTREE_ROOT",
  "use_new_terminal": false,
  "allow_concurrent_runs": true,
  "reveal": "on_error"
}

**Development Server:**

{
  "label": "Start dev server",
  "command": "deno task dev",
  "cwd": "$ZED_WORKTREE_ROOT",
  "use_new_terminal": true,
  "allow_concurrent_runs": false,
  "reveal": "always"
}

**Formatter:**

{
  "label": "Format file",
  "command": "deno fmt $ZED_FILE",
  "cwd": "$ZED_WORKTREE_ROOT",
  "use_new_terminal": false,
  "allow_concurrent_runs": true,
  "reveal": "never"
}