A collection of configurations, prompts, and tools that I use to enhance my AI-assisted development workflow.
Custom configurations for Claude Code.
Custom status line that provides real-time context awareness and project visibility.
The custom status line provides at-a-glance information about your development session:
- Project name - Current working directory
- Git status - Branch name and working tree state (clean ✓ or dirty ●)
- Model information - Currently active Claude model (Sonnet 4, Opus, etc.)
- Context usage - Token usage with intelligent buffer calculations and color-coded warnings
The status line uses color coding to communicate state:
- Green git indicator = clean working tree
- Red git indicator = uncommitted changes
- Gray context indicator = normal usage (<90%)
- Yellow context indicator = approaching auto-compact (90-94%)
- Red context indicator = auto-compact imminent (≥95%)
- Copy settings configuration:
cp claude/settings.json ~/.claude/settings.json
Note: Ensure that the
"command"path inclaude/settings.jsonmatches the location where you'll installstatusline-command.sh.
-
Install the status line script:
cp claude/statusline-command.sh ~/.claude/statusline-command.sh chmod +x ~/.claude/statusline-command.sh
-
Restart Claude Code to apply the new configuration.
settings.json
- Enables the custom status line command (ensure that the path to the statusline-command.sh is correct)
- Activates extended thinking mode for better reasoning
statusline-command.sh
- Parses Claude Code's session data (workspace, model, transcript)
- Calculates real-time context usage from the transcript JSONL file
- Includes a 22.5% auto-compact buffer to match
/contextcommand behavior - Renders a powerline-style status bar with ANSI color codes
- Integrates git status checking for branch and working tree state
The context calculation accounts for:
- Input tokens
- Cache read tokens
- Cache creation tokens
- Auto-compact buffer (~45k tokens for 200k limit)
This gives you an accurate representation of when Claude Code will trigger auto-compact, helping you manage context proactively.
Curated prompts and reference documents for AI-assisted development.
File: prompts/design/design_principles.md
A comprehensive reference document covering fundamental design and accessibility principles for building high-quality software interfaces.
This document consolidates four essential frameworks:
-
Dieter Rams' 10 Principles of Good Design
- Originally for physical products, adapted for software
- Emphasizes innovation, utility, aesthetics, clarity, and minimalism
- Includes practices, watch-outs, and team checklists
-
Ben Shneiderman's Eight Golden Rules of Interface Design
- Focus on consistency, shortcuts, feedback, and error handling
- Practical heuristics for design reviews
- Instrumentation and quality gate recommendations
-
Nielsen's 10 Usability Heuristics
- System visibility, user control, error prevention
- Recognition over recall, flexibility and efficiency
- Includes design review checklists and metrics guidance
-
WCAG 2.0 Accessibility Guidelines
- POUR principles (Perceivable, Operable, Understandable, Robust)
- AA conformance targets with concrete examples
- Engineering-friendly acceptance criteria and testing strategies
MIT - Feel free to use, modify, and share.
