Skills Are Structured Expertise
Most AI coding tools give you prompts. SpecWeave gives you structured expertise.
A skill is a SKILL.md file with instructions — Claude adds it to its toolkit, uses it automatically when relevant, or you invoke it directly with /skill-name. Each skill packages domain knowledge: patterns, rules, and examples that make AI produce consistent, production-grade output instead of generic guesses. It runs the same way every time — but you can customize it without touching the source.
What Makes a Skill Different from a Prompt?
| Prompt | SpecWeave Skill | |
|---|---|---|
| Reusable | Copy-paste across sessions | Invoke by name: sw:grill |
| Extensible | Edit the original text | Defined in SKILL.md — override via skill-memories/ without touching the source |
| Composable | Manual chaining | Skills invoke other skills: PM → Architect → Frontend |
| Stateful | Forgets everything | Learns from corrections permanently |
| Testable | "Did it work?" | Quality gates verify output automatically |
The Power: You Shape AI Behavior
Instead of writing this every session:
"When generating React components, always use React Hook Form with Zod
validation, Tailwind for styling, import from @/components/ui, and
write Vitest tests with Testing Library..."
You write it once in a SKILL.md file — the skill definition itself:
# SKILL.md
---
name: frontend-design
description: Create production-grade frontend interfaces
triggers:
- build web component
- create page
- design UI
---
## Rules
### Form Handling
- Use React Hook Form for all forms
- Combine with Zod for validation schemas
- Never use plain useState for form state
### Styling
- Tailwind utilities only, no inline styles
- Import from @/components/ui design system
Next session, next agent, next month — the AI already knows your patterns.
Want to customize a skill without editing the original? Use skill memories — they override defaults while the core skill keeps getting updates:
# .specweave/skill-memories/frontend.md
### Styling Override
- Use CSS Modules instead of Tailwind
- Import from @/styles design tokens
44 Core Skills Cover the Full Lifecycle
SpecWeave ships with 44 core skills for planning, execution, quality, and sync:
- PM (
sw:pm) — writes user stories with acceptance criteria - Architect (
sw:architect) — designs systems, writes ADRs - Code Review (
sw:code-reviewer) — parallel multi-agent review with confidence scoring - Testing (
sw:e2e,sw:tdd-cycle) — E2E automation, TDD red-green-refactor - Quality (
sw:grill,sw:judge-llm) — critical code review and LLM-as-Judge validation - Debug (
sw:debug) — systematic 4-phase debugging with escalation - Team Lead (
sw:team-lead) — parallel multi-agent orchestration - Release (
sw:release-expert,sw:npm) — multi-repo release coordination
Need domain-specific skills like frontend frameworks, backend languages, or security? Install them from the vskill marketplace — they plug into SpecWeave's core workflow without modification.
Each skill can be customized independently. Your frontend preferences don't affect your backend patterns.
Zero Learning Curve
Skills package expertise so you can focus on what to build, not how to configure your tools:
- No workflow setup — skills handle planning, architecture, and testing automatically
- No context management — specs persist across sessions in version-controlled files
- No manual documentation — living docs generate and stay current as you build
- No tool configuration — skills auto-activate based on your project's tech stack
Install SpecWeave, describe your feature, and skills do the rest.
npm install -g specweave
specweave init .
Extensibility: Stable Core, Customizable Extensions
Skills follow a stable core with customizable extensions:
- Stable core — don't edit SKILL.md files; they keep getting updates
- Your extensions — customize via
.specweave/skill-memories/*.md
Your customizations override defaults. Original skills keep getting updates. No fork needed.