Agents & Workflows — Delegate at Scale

# --print runs Claude non-interactively, outputs result, then exits
claude --print "Review all .js files in src/ for unused variables and write a report to review.md"

# Run it in the background (& sends it to the background)
claude --print "Audit package.json for outdated dependencies and create a report" &

# Pipe output to a file
claude --print "Generate a weekly summary of changes in this repo" > weekly-summary.md

# Run with a specific model for cost control
claude --print --model claude-haiku-4-5 "Format all markdown files in docs/"

# Write a detailed task file
cat > agent-task.md << 'EOF'
Review the pull request diff below and write a code review.
Focus on: security issues, performance problems, missing tests.
Format: numbered list, one issue per line.
EOF

# Run the agent with the task file as input
claude --print --input-file agent-task.md > code-review.md

{
  "hooks": {
    // Run prettier after every file edit
    "PostToolUse": [
      {
        "matcher": "Edit",
        "hooks": [
          { "type": "command", "command": "prettier --write $CLAUDE_TOOL_INPUT_PATH" }
        ]
      }
    ],
    // Send a desktop notification when Claude finishes
    "Stop": [
      {
        "hooks": [
          { "type": "command", "command": "osascript -e 'display notification \"Claude finished\" with title \"Claude Code\"'" }
        ]
      }
    ]
  }
}

# Auto-run tests after every file edit
"command": "npm test --silent"

# Auto-lint Python files after edits
"command": "ruff check $CLAUDE_TOOL_INPUT_PATH --fix"

# Git commit automatically when Claude stops
"command": "git add -A && git commit -m 'claude: auto-commit'"

# Mac notification when Claude finishes
"command": "osascript -e 'display notification \"Done\" with title \"Claude Code\"'"

# Windows notification
"command": "powershell -command \"[System.Windows.Forms.MessageBox]::Show('Claude finished')\""

# Linux notification
"command": "notify-send 'Claude Code' 'Task complete'"

# Schedule a one-time task
Use the schedule skill to run this task tomorrow at 9am:
"Read all open Linear issues, write a prioritized daily plan
to daily-plan.md, and create a Notion page from it."

# Schedule a recurring weekly task
Use the schedule skill to run every Monday at 8am:
"Read last week's commits with git log, summarize what changed,
and create a weekly-update.md in the project root."

# Schedule a daily standup prep
Use the schedule skill to run every weekday at 8:30am:
"Read my Linear issues and Notion meeting notes from yesterday,
then write a standup summary to standup.md."

# Open your crontab file
crontab -e

# Add a line in this format: minute hour day month weekday command
# Run every Monday at 9:00 AM
0 9 * * 1 cd ~/my-project && claude --print "Generate weekly summary" >> weekly.log 2>&1

# Run every day at 8:30 AM (weekdays only)
30 8 * * 1-5 cd ~/my-project && claude --print "Write standup summary to standup.md"

REM Open Task Scheduler → Create Basic Task
REM Set trigger: Daily at 9:00 AM
REM Set action: Start a program
REM Program: cmd.exe
REM Arguments:
/c "cd %USERPROFILE%\my-project && claude --print "Generate weekly summary" >> weekly.log"

You are an orchestrator. Break this task into 3 parallel subtasks
and execute each one completely before combining the results.

Task: Audit this codebase for quality issues.

Subtask 1: Scan all .js files for unused variables and imports.
Subtask 2: Check all API endpoints for missing input validation.
Subtask 3: Review package.json for outdated or vulnerable dependencies.

When all three are done, write a combined report to audit-report.md
with a section for each subtask and a priority-ranked action list.

# In settings.json Stop hook:
"command": "git add -A && git commit -m \"$(claude --print 'Write a one-line commit message for these staged changes')\" && gh pr create --fill"

0 23 * * 1-5 cd ~/my-project && claude --print "Review today's git commits with git log --since=today, identify any risky changes, and write a nightly-review.md"

After every file edit, run npm test. If tests fail, read the
failure output and fix the code. Keep iterating until all
tests pass. Don't ask me for confirmation between iterations.

Read: git log for this week's commits, open Linear issues,
and any Notion pages updated this week.

Write a weekly-report.md with:
- What shipped (from git log)
- What's in progress (from Linear)
- Blockers (from Linear high-priority issues)
- Next week's focus (top 3 Linear priorities)

Keep it under 300 words. Executive summary tone.