Copy/Paste Oneliners

No explanation needed. Just copy, paste, and run.

Before You Commit

# Review only what you changed
unfault review --uncommitted

# Same thing, detailed output
unfault review --uncommitted --output full

# JSON for scripts
unfault review --uncommitted --output json

For Coding Agents

Optimized output for Claude, Cursor, Copilot, and other AI assistants.

# LLM-optimized JSON — compact, actionable, no noise
unfault review --llm

# Focus on your changes only
unfault review --uncommitted --llm

# Specific files with full context
unfault review --llm --file src/api/routes.py --file src/api/handlers.py

# Limit to top 20 most severe findings
unfault review --llm --top 20

# The full combo: uncommitted, LLM format, top issues
unfault review --uncommitted --llm --top 10

Paste Into Your Agent

Add this to your AGENTS.md or system prompt:

Before committing, run:
\`\`\`bash
unfault review --uncommitted --llm
\`\`\`

If findings are reported, fix high-severity issues before committing.
Use the `files_allowed_to_change` field to scope your edits.

CI/CD Pipelines

GitHub Actions

# SARIF for GitHub Code Scanning
unfault review --output sarif > results.sarif

- name: Run Unfault
  run: unfault review --output sarif > results.sarif

- name: Upload SARIF
  uses: github/codeql-action/upload-sarif@v3
  with:
    sarif_file: results.sarif

Exit Code Checks

# Fail if findings detected (exit 5)
unfault review && echo "Clean!" || echo "Findings found"

# Explicit check
unfault review --output json
if [ $? -eq 5 ]; then
  echo "Review the findings above"
  exit 1
fi

JSON for Custom Processing

# Pipe to jq for filtering
unfault review --output json | jq '.findings | map(select(.severity == "High"))'

# Count findings
unfault review --output json | jq '.findings | length'

Deep Analysis

# Full detailed output with suggested fixes
unfault review --output full

# Focus on one dimension
unfault review --dimension stability --output full
unfault review --dimension performance --output full
unfault review --dimension correctness --output full

# Include test files (excluded by default)
unfault review --include-tests --output full

# Preview fixes without applying
unfault review --dry-run

Impact Analysis

# What breaks if I change this file?
unfault graph impact src/core/auth.py

# Find the most connected files
unfault graph critical --limit 10

# What uses this library?
unfault graph library requests
unfault graph library httpx

# External dependencies of a file
unfault graph deps src/api/client.py

# Graph statistics
unfault graph stats

Ask Your Codebase

# Quick semantic search
unfault ask "functions without error handling"
unfault ask "routes without authentication"
unfault ask "database queries without timeouts"

# With LLM-powered synthesis
unfault ask "how do we handle retries?" --llm
unfault ask "what happens if the database goes down?" --llm

Configuration

# Check auth status
unfault status

# Configure LLM for `ask --llm`
unfault config llm openai --model gpt-4o
unfault config llm anthropic --model claude-3-5-sonnet-latest
unfault config llm ollama --model llama3.2

# View current config
unfault config show
unfault config llm show

Quick Reference

Flag	What it does
`--uncommitted`	Only findings for uncommitted files
`--file PATH`	Only findings for specific files (repeatable)
`--llm`	Compact JSON optimized for AI agents
`--top N`	Limit findings (default 50, max 200)
`--output full`	Detailed output with fix suggestions
`--output json`	Machine-readable JSON
`--output sarif`	SARIF 2.1.0 for GitHub/IDE integration
`--dimension X`	Focus on one dimension
`--include-tests`	Include test files in analysis
`--dry-run`	Preview fixes without applying

Exit Codes

Code	Meaning
0	Success, no findings
5	Success, findings detected
1	General error
2	Config error
3	Auth failed
4	Network error