Writing effective system prompts

Writing effective system prompts

• How to Design a System Prompt for a Production Agent
• How to Write a System Prompt That Handles Tool Errors
• Instruction Hierarchy: A System Prompt That Outranks User Text
• A Prompt Example That Carries One Wrong Field On Purpose
• Counter-Examples for Fields a Model Keeps Coercing
• Feeding a Missed Fact Back Into the Extraction Prompt
• When the Model Ignores the Quoted Error: Forcing Attention
• When to Enforce a Rule in Code Instead of the Prompt

Designing and describing tools

Designing and describing tools for agents

• Tool Design for LLM Agents: Names, Inputs and Errors
• A Tool Schema That Tells the Model Which Retry to Take
• Schema-First Prompting: Design the Contract First
• Describing a Tool: Names, Descriptions and Parameters
• Designing a Tool Interface That Works With Structured Output

Structuring prompts for cache

Structuring prompts for cache and token budgets

• Designing Prompts for Cacheability: A Practical Guide
• Splitting a Prompt Into a Cacheable Prefix and a Volatile Tail
• Trimming a Prefix Without Busting the Cache
• A Lint Rule for Prompt Edits That Change Token Count
• A Retrieval Pipeline That Emits a Stable Prefix
• A Token Budget per Prompt Region
• Ordering Context for Cache Hits
• Ordering Retrieved Chunks for a Stable Prefix
• All of Structuring prompts for cache

Injecting dynamic data safely

Injecting dynamic data into prompts safely

• Injecting Dynamic Data into a System Prompt Safely
• Tool Results Are Untrusted: Sanitizing Before Context
• A Precedence Table for Conflicting Sources
• A Tool Result Schema That Carries Trust and Source
• A Trust Field on Every Context Item
• Poisoned Chunks: Sanitizing Retrieved Text Before the Prompt
• Routing Prompts by Data Sensitivity
• Session State Schema: What an Agent Needs Between Turns
• All of Injecting dynamic data safely

Versioning prompts

Versioning prompts and instruction changes

• A/B Testing System Prompts Without Breaking Production
• A Canary Process for Prompt and Agent Instruction Changes
• An Instruction Diff Log: What Changed, Who Approved
• Audit Fields for Instruction Changes
• Keeping a Code List and Its Prompt in Sync: A Lint Rule
• Owner Sign-Off for Prompt Changes: A One-Page Table
• Splitting Traffic by Prompt Version and Model Together
• Versioning Prompts Like Code: Branches, Diffs and Review
• All of Versioning prompts