Claude Code Quick Reference

Quick lookup for common commands and workflows.

Comprehensive Guide

For detailed explanations, see Working with Claude Code

Logsift Commands

Run with Explicit Command

/logsift "bash ~/dotfiles/tests/install/test-install-wsl-docker.sh --reuse" 15
  • First argument: Exact command in quotes
  • Second argument: Timeout in minutes (optional, default: 10)

Run with Natural Language

/logsift-auto run wsl docker test with reuse flag, 15 minutes
  • Describe what you want to run
  • Claude figures out the exact command

Built-in Commands

/cost          # View token usage this session
/clear         # Clear context between work sessions
/model         # View/change active model
/help          # View available commands

Commit Agent

# Invoke with natural language
"Let's commit this work"
"Create a commit for these changes"
"Commit the staged files"

# Agent workflow:
# 1. Analyzes staged changes
# 2. Groups into atomic commits
# 3. Generates conventional commit messages
# 4. Runs pre-commit (background → logsift)
# 5. Fixes errors iteratively
# 6. Reports summary

# Benefits:
# - Saves ~5000-6000 tokens per commit
# - Isolates commit workflow from main context
# - Handles pre-commit automation
# - Splits multi-concern changes intelligently

Analysis Tools

# View metrics summary
analyze-claude-metrics

# Detailed breakdown
analyze-claude-metrics --details

# Specific date
analyze-claude-metrics --date 2025-12-03

Common Workflows

Run and Fix Script

# 1. Run with logsift
/logsift "bash script.sh" 15

# 2. Claude analyzes errors and fixes them

# 3. Claude re-runs automatically

# 4. Iterate until all errors resolved

Track Quality Manually

After significant sessions, add to .claude/metrics/quality-log.md:

## YYYY-MM-DD HH:MM - Session ID

**Command**: `/logsift "command"`
**Quantitative**: Errors X → 0, Iterations Y, Tokens Z
**Qualitative**: Correctness ✅, Efficiency ✅, Methodology ✅
**Notes**: What worked well, what could improve

Check Token Usage

# During session
/cost

# Enable detailed tracking
export CLAUDE_CODE_ENABLE_TELEMETRY=1

Error Fixing Phases

  1. Initial Analysis: Read all errors, identify patterns
  2. Root Cause Investigation: Related errors? Independent errors?
  3. Solution Strategy: Fix root cause or fix independently
  4. Iterative Fix-and-Rerun: Verify fixes, continue until resolved
  5. Verification: Confirm robustness

Quick Decisions

Use /logsift when:

  • You know the exact command
  • Fast execution, no ambiguity

Use /logsift-auto when:

  • Describing what to run
  • Don't know exact paths/flags

Fix as root cause when:

  • Same file/module/dependency
  • Error messages indicate same issue

Fix independently when:

  • Different scripts/components
  • Unrelated error types

Key Principles

✅ Do:

  • Read files before editing
  • Determine error relationships
  • Fix root causes when they exist
  • Fix independently when appropriate
  • Use context for quality fixes

❌ Don't:

  • Background logsift processes
  • Suppress errors without understanding
  • Guess and check without reading files
  • Force false root cause connections
  • Stop after first error resolved

Troubleshooting

# Logsift not found
cargo install logsift

# Metrics not tracking
chmod +x .claude/hooks/track-command-metrics

# Need more details from logsift
cat ~/.local/share/logsift/logs/latest-session.json

File Locations

.claude/
├── commands/
│   ├── logsift.md           # /logsift definition
│   └── logsift-auto.md      # /logsift-auto definition
├── metrics/
│   ├── README.md            # Metrics framework
│   ├── quality-log.md       # Manual quality tracking
│   └── command-metrics-*.jsonl  # Automated logs
├── hooks/
│   └── track-command-metrics    # Metrics collection
└── settings.json            # Hook configuration

apps/common/
└── analyze-claude-metrics  # Analysis tool