Prompt Design Patterns Collection

Reusable prompt structures organized by purpose for AI agents, daily coding, and document generation.

Classification View

Category	Purpose	Typical Triggers	Quality Metrics	Risks
Framing	Context boundary fixation	"You are", scenario declaration	Reproducibility, deviation rate	Role injection
Structured Output	Machine readability	JSON/YAML specification, schema examples	Parse success rate	Schema drift
Thought Expansion	Reasoning visualization	Chain-of-Thought / Scratchpad	Solution transparency	Verbosity, information leakage
Constraint/Validation	Guardrails	Failure condition enumeration, rule violation warnings	Violation detection rate	Over-constraint reduces emergence
Reduction/Extraction	Summarization/normalization	Tagged extraction targets	Omission rate, redundancy rate	Partial context loss

Basic Template (Skeleton)

<ROLE>
Objective: <WHAT>
Preconditions (constrained facts): <FACTS>
Input format:
<FORMAT_SPEC>
Expected output (JSON Schema excerpt): <SCHEMA_SNIPPET>
Quality checks:
- Validity: <CRITERIA>
- Safety: <CRITERIA>
Procedure:
1. Problem restatement
2. Verifiable decomposition (bullet points)
3. Output generation (specified format)
4. Rule check & retry conditions

Anti-patterns

Name	Symptoms	Impact	Improvement
Directionless enumeration	Scattered requirements	Output inconsistency	Separate and structure objectives/constraints
Excessive few-shot	Lengthy samples	Context overflow	Compression/pattern abstraction
Instruction conflict	Contradictory commands mixed	Rising deviation rate	Explicit priority table

Next Actions

• JSON Schema generation automation macro integration
• Evaluation metrics (reproducibility/structuring rate) scripting

Back: index.md