Prompt engineering formulates instructions and context to guide an AI model’s output. A prompt that leaves the model to guess where a change belongs may work on a small codebase, but can miss important work. Rephrasing alone is not a substitute for identifying the missing context or checking the result. section 2.4.3
Make the missing condition explicit
A test-migration prompt asking to rework non-JUnit 5 tests specifies the framework but not where to look. A more targeted prompt asks whether tests import JBoss-related libraries and, if so, to replace those imports with JBoss-agnostic code using Jupiter. Both approaches produced similar changes in this case, but the second states the condition that exposed the missed step. section 2.4.3
For generated SQL, a prompt can combine a role, the user’s request, table creation statements, and recent queries. This gives the model concrete schema and usage context rather than relying on a request alone. The resulting SQL still needs evaluation; framing the model as an expert is not evidence that its output is correct. section 6.2
Listing 6.11 The SqlGeneratorServiceImpl#buildPrompt() method
String buildPrompt(SqlGeneratorQueryRequest request) {
StringBuilder prompt = new StringBuilder();
prompt.append("You are a SQL expert. Generate a SQL query based on the following information:\n\n"); #A
prompt.append("User Request: ").append(request.getUserPrompt()).append("\n\n"); #B
prompt.append("Available Tables:\n"); #C
for (Table table : request.getTables()) {
prompt.append("- ").append(table.getCreateTableSql()).append("\n"); #C
}
#D
if (request.getNLastQueries() != null && !request.getNLastQueries().isEmpty()) {
prompt.append("\nRecent Queries for those tables:\n");
for (int i = 0; i < request.getNLastQueries().size(); i++) {
prompt.append(i + 1).append(". ").append(request.getNLastQueries().get(i)).append("\n");
}
prompt.append("\n");
}
prompt.append("Please generate a valid SQL query that addresses the user's request. ");
prompt.append("Only return the SQL query without any additional explanation or formatting."); #E
return prompt.toString();
}Inspect how the SQL prompt is assembled from the request, table definitions, and recent queries. section 6.2
Check the assumption inside the prompt
A request to move away from a single-threaded implementation led to several concurrent strategies without a challenge to whether concurrency was needed. The generated benchmark also assumed 100,000 accounts; a production-like scenario used 10,000. Its custom benchmark had potential measurement problems, so the results should not be treated as conclusive without a sounder benchmark such as JMH. section 7.1.2
Use prompts as part of an engineering loop:
- State the task, relevant scope, and constraints; inspect what the model changed and refine the prompt when a step is missed. section 2.4.3
- Question constraints before asking for an implementation, and validate performance claims with representative inputs and appropriate benchmarks. section 7.1.2
- Keep reusable rules focused and current. Record useful discoveries, and periodically check rules for unused, duplicated, or contradictory guidance. section 9.1.3
- Match effort to the problem: a direct prompt may retrieve a known answer, while other tasks require reasoning and checks. AI review can identify issues, but cannot infer unstated requirements. section 3.6.4 section 8.2.3


















































