Reflection, writing, and design skills for Claude Code.
curl -fsSL https://raw.githubusercontent.com/tslateman/duet/main/install.sh | bash
Or clone directly:
git clone https://github.com/tslateman/duet.git ~/.claude/plugins/duet
git -C ~/.claude/plugins/duet pull
Or run the install script again; it pulls if already installed.
Two voices in dialogue, neither dominant. The composition emerges from the interaction. Read more.
You don't need to memorize the catalog. Describe what you're doing in plain language and Claude routes to the right skill automatically:
- "debug this" →
/debugging - "review my PR" →
/review - "how should I test this" →
/testing
Explore the full catalog by problem type in Skills by Concern.
Use /install-statusline to set up a lean status bar for Claude Code sessions.
- Skills auto-invoke from context and teach methodology
- Commands are workflows you invoke by name
- Agents are subprocesses for parallel work
| Command | Description |
|---|---|
/capture-memory |
Capture learnings to MEMORY.md, output compact summary |
/debrief |
Review what agents accomplished, surface loose ends across sessions |
/lint |
Run markdownlint on specified files or the whole project |
/memory-audit |
Audit project memory files for staleness, orphans, bloat, and duplication |
/ponder [n] |
Surface a question worth sitting with |
/probe |
Probe recent changes with 2-3 pointed questions |
/retro |
Reflect on recent work—capture learnings and surface what to think about next |
/sweep |
Post-op check for artifacts, damage, and stale references after agent work |
/vamp |
Survey the project and choose what to play next |
| Skill | Description |
|---|---|
/adr |
Generate Architecture Decision Records that capture the reasoning behind technical decisions. |
/debugging |
Systematic debugging that applies the scientific method to failures. |
/design |
Intentional design thinking for interfaces, components, and systems. |
/excalidraw |
Create Excalidraw diagrams programmatically. |
/ia |
Evaluate and improve how information is organized for findability. |
/mermaid |
Create diagrams and visualizations using Mermaid.js v11 syntax. |
/naming |
Evaluate and improve names in code using naming as a design diagnostic. |
/performance |
Profiling methodology and optimization strategy for performance work. |
/prose |
Apply Strunk's writing rules to prose—documentation, commits, error messages, UI text. Clearer, stronger, more professional. Also handles sharpening and tightening drafts by subtraction. |
/research |
Systematic technical investigation—evidence gathering, option comparison, and actionable recommendations. |
/review |
Framework for code review that captures context future maintainers need—concerns raised, alternatives rejected, risks accepted. |
/testing |
Design test strategy using Beck's Test Desiderata — which properties matter, which tradeoffs to make. |
| Agent | Description |
|---|---|
code-simplifier |
Simplifies and refines code for clarity, consistency, and maintainability while preserving all functionality. Focuses on recently modified code unless instructed otherwise. |
Seven skills include RECIPE.md files that teach Bach (a multi-agent
orchestrator) how to decompose domain tasks into parallel workers. Each recipe
defines roles, inputs, and coordination patterns for multi-agent work. See
FRAMEWORKS.md for the full list.
duet/
├── .claude-plugin/
│ └── plugin.json # Plugin metadata
├── commands/ # Slash commands
├── skills/ # Skills with supporting files
│ └── <name>/SKILL.md
├── agents/ # Subagent definitions
├── statusline/ # Status bar script
├── scripts/ # Build and maintenance scripts
└── notes/ # Project thinking
- Claude Code
- Git
- GitHub CLI (
gh)
MIT