Tips & Tricks

# Add to ~/.bashrc or ~/.zshrc
alias ai='openclaw run'
alias aif='openclaw run --file'
alias ais='openclaw run --stdin'

# Usage
ai "What is the capital of France?"
aif report.csv "Analyse this CSV"
cat error.log | ais "Find the root cause"

# JSON output for piping to jq
openclaw run "Extract name, email from this text as JSON" --file contacts.txt --format json | jq '.[]'

# Markdown for documents
openclaw run "Write a README for this project" --file src/ --format markdown --output README.md

# Plain text for scripts
openclaw run "Should I buy or sell? Reply: buy, sell, or hold" --file stock-data.csv --format text

# Use a different model for just one command
OPENCLAW_MODEL=gpt-4o openclaw run "Complex analysis task" --file big-doc.pdf

# Temporarily increase timeout
OPENCLAW_TIMEOUT=180 openclaw run "Very long task" --file huge-file.txt

# Use a different config file
openclaw run "task" --config ~/configs/local.yaml

# Analyse clipboard contents (macOS)
pbpaste | openclaw run "Check this text for grammar errors"

# Analyse git diff before committing
git diff --staged | openclaw run "Review this diff for bugs and suggest a commit message"

# Process command output
df -h | openclaw run "Which disk is most full? Should I be worried?"

# Switch between work (expensive cloud) and home (free local) profiles
openclaw run "task" --profile work    # uses gpt-4o
openclaw run "task" --profile home    # uses ollama/llama3.2

# ~/.openclaw/config.yaml
profiles:
  work:
    provider: openai
    model: gpt-4o
  home:
    provider: ollama
    model: llama3.2

# Pipe output of one task to the next
git log --oneline --since="7 days ago" | \
  openclaw run "Convert these commits to a user-facing changelog" | \
  openclaw run "Translate the changelog to Spanish"

# Pass multiple files and merge results
cat report-1.txt report-2.txt | openclaw run "Summarize all findings"

# Store intermediate results
openclaw run "Extract action items" --file meeting.md > actions.txt
openclaw run "Prioritize by urgency and assign estimates" --file actions.txt

# See the full prompt sent to the AI
openclaw run "Analyze this" --file data.txt --debug

# Verbose output: token count, latency, model
openclaw run "Hello!" --verbose

# Dry run: print what OpenClaw would do without executing
openclaw run "Delete all temp files" --dry-run

# Log all API calls to a file
openclaw run "Process report" --log /tmp/openclaw-debug.log

# Set a system prompt globally
openclaw config set system_prompt "You are a senior Python developer. Always use type hints, write docstrings, and include error handling in every function."

# Override system prompt for a single run
openclaw run "Write a URL parser" \
  --system "You are a security-focused developer. Always validate input and sanitize URLs."

# Use a system prompt file
openclaw run "Review this code" \
  --system-file ./prompts/code-reviewer.txt \
  --file main.py

# Process all files in a directory
for f in ./reports/*.pdf; do
  echo "Processing $f..."
  openclaw run "Extract: title, date, key findings, recommendations" \
    --file "$f" \
    --format json \
    --output "./summaries/$(basename $f .pdf).json"
done

# Parallel processing (faster, uses more API quota)
ls *.md | xargs -P4 -I{} openclaw run "Summarize this document" --file {} --output {}.summary.txt

# Quick Telegram notification from a script
openclaw telegram send "Backup complete: $(du -sh /backups | cut -f1)"

# Git-aware automation
git diff HEAD~1 | openclaw run "Summarize what changed in this diff"

# Clipboard integration (macOS)
pbpaste | openclaw run "Fix grammar and rephrase professionally" | pbcopy

# Clipboard integration (Linux, requires xclip)
xclip -o | openclaw run "Translate to French" | xclip -sel clip

# Quick web page summary (using curl)
curl -s https://example.com | openclaw run --stdin "Summarize this webpage"

# prompts/code-review.txt
Review the following code as a senior engineer:
1. Flag any bugs or logic errors
2. Identify security vulnerabilities
3. Suggest performance improvements
4. Check adherence to clean code principles
Format output as a numbered checklist.

# Use the template
openclaw run --prompt-file prompts/code-review.txt --file main.py

# Combine template + extra instruction
openclaw run --prompt-file prompts/code-review.txt --append "Focus on async safety" --file async_handler.py