prompt-builder

The prompt-builder skill is designed to generate complete, deterministic prompts for BAZINGA agents through a structured Python script workflow. It addresses the challenge of consistently creating fully-formed agent prompts that integrate context, specialization, and agent-specific definitions without manual intervention. By automating this process, it ensures that agents like Developers, QA Experts, Tech Leads, PMs, and other roles receive prompts that are accurate, validated, and ready for immediate execution.

The skill's core capabilities include reading task specializations, context packages, reasoning patterns, and full agent definitions from both the database and the filesystem. It applies model-specific token budgets, validates required markers, and outputs both the prompt file and a JSON summary detailing success status, token usage, line count, and marker validation results. It also supports additional fields for retries, compact return protocols, and PM-specific spawns, making it versatile for different orchestrator needs.

This skill is primarily invoked right before spawning any BAZINGA agent, ensuring that agents operate with the most up-to-date context and parameters. Target users include orchestrators and developers within the BAZINGA ecosystem who need reliable prompt generation, as well as QA and PM teams who require validated prompts for testing and task handoffs. Its deterministic nature and integration with database and filesystem sources make it essential for automated agent orchestration in multi-role workflows.