nWave-ai/nWave / nWave/skills/nw-command-design-patterns-reduction
nw-command-design-patterns-reduction
UnverifiedWhat is reducible in a bloated command - the duplication triangle, the anti-pattern catalog, and the compress/never-compress rules
Command Reduction: Duplication, Anti-Patterns, Compression (KNOWLEDGE)
Kind: KNOWLEDGE (reference). Composed by nw-command-design-patterns core. Fires when: a command is bloated and the question is "what here is reducible / what should be removed / how far can I compress" (the duplication triangle + anti-pattern + compression rules). No forced sequence.
The Duplication Triangle
Commands duplicate content in three directions, all waste tokens:
- Command-to-Command: Orchestrator briefings, agent registries, parameter parsing repeated in 5-12 files (~620 lines waste)
- Command-to-Agent: Domain knowledge belonging in agents (~1,300 lines waste). Examples: TDD phases in execute.md, DIVIO templates in document.md, refactoring hierarchies in refactor.md
- Command-to-Self: develop.md embeds other commands inline (~1,000 lines)
Fix: Extract shared content to preamble skill. Move domain knowledge to agents. Have orchestrators reference sub-commands.
Anti-Patterns
| Anti-pattern | Impact | Fix |
|---|---|---|
| Procedural overload | Step-by-step for capable agents wastes tokens, "lost in the middle" | Declare goal + constraints, let agent apply methodology |
| Duplicated briefings | Same orchestrator constraints in every command (30-80 lines each) | Extract to shared preamble, reference once |
| Embedded domain knowledge | Refactoring hierarchies, review criteria, TDD cycles in commands | Move to agent definitions or skills |
| Aggressive language | "CRITICAL/MANDATORY/MUST" causes overtriggering in Opus 4.6 | Direct statements without emphasis markers |
| Example overload | 50+ lines of JSON examples | 2-3 canonical examples suffice |
| Inline validation logic | Prompt template validation in command text | Platform/hook responsibility |
| Dead code | Deprecated formats, aspirational metrics, old signatures | Remove; version control preserves history |
| Verbose JSON state examples | 200+ lines of unused JSON | Show actual format (pipe-delimited), 3 examples max |
Compression Guidelines
When optimizing command files for token efficiency:
Safe to compress:
- Prose descriptions → pipe-delimited
- Verbose explanations → imperative voice
- Filler words ("in order to", "it is important to") → remove
- Related bullet items → single line with
|separators
Never compress:
### Example N:section headers — keep verbatim (eval tools and agents depend on these)- AskUserQuestion decision tree options — these are runtime menu items, not documentation
**Question**:lines in decision points — runtime behavior- Code blocks and YAML — preserve verbatim
- YAML frontmatter — preserve exactly
Compression evidence: Pipe-delimited compression achieves 15-30% token reduction on prose-heavy files. Code-heavy files (PBT skills, code examples) yield <5%. Average across framework: ~7.4% overall.
Orchestrator skill loading section: Commands dispatching sub-agents must include SKILL_LOADING in the Task prompt reminding the agent to read its skills at ~/.claude/skills/nw-{skill-name}/SKILL.md. Sub-agents can invoke the Skill tool (but not slash commands), though a skill carrying disable-model-invocation: true stays unreachable through it and must be READ at the path above; the skills: frontmatter field eagerly preloads full skill content into custom subagents, so this reminder targets skills meant to load at point-of-use instead. Without it, sub-agents operate without domain knowledge.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 195k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 195k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 195k | Updated | Claude CodeCodex | Verified | Install |
| web-artifacts-builderanthropics/skillsSuite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn/ui). Use for complex artifacts requiring state management, routing, or shadcn/ui components - not for simple single-file HTML/JSX artifacts. | Stars: 195k | Updated | Claude CodeCodex | Unverified | Install |