If you're building LLM pipelines and tired of prompt spaghetti, this skill teaches you DSPy 3.2.x the right way: typed Signatures for I/O contracts, dspy.Module subclasses for composition, and predictors like ChainOfThought and ReAct instead of raw prompt strings. It's opinionated about modern idioms (no more TypedPredictor, no legacy dspy.OpenAI calls) and includes a ten-point anti-pattern list that will save you from common mistakes. The canonical template and predictor cheatsheet are copy-paste ready. Use this when starting any DSPy project or refactoring prompt-heavy code into something you can actually serialize, version, and optimize programmatically.
npx -y skills add intertwine/dspy-agent-skills --skill dspy-fundamentals --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
mindrally/skills
giuseppe-trisciuoglio/developer-kit
syncfusion/react-ui-components-skills
supercent-io/skills-template
binjuhor/shadcn-lar