steering-specs-generator

The Steering Specs Generator is a powerful tool designed to bridge the gap between human expertise and machine-readable documentation. It excels at capturing the tribal, often unspoken, engineering knowledge that lives within development teams. By conducting structured, context-aware interviews, this agent extracts essential domain and architectural insights—ranging from design invariants and business contracts to codebase topology—and transforms them into high-quality, actionable documentation known as steerings. Designed for architects, lead engineers, and team leads, it ensures that your team’s unique practices, conventions, and quality standards are codified for both human developers and future AI agents to follow accurately.

• Automated discovery of existing project conventions, README files, and architecture documents through intelligent repo scanning.
• Eight distinct interview packs covering codebase topology, business domain contracts, security, testing strategies, and more.
• Support for custom topics, allowing teams to document specific, niche workflows or internal library constraints.
• Interactive and fast-mode interview capabilities to suit varying levels of user engagement and time availability.
• Produces standardized, markdown-formatted outputs following an Intent-Rules-Practices-Meta structure for maximum clarity.
• Intelligent session management, allowing users to pause, save, and resume knowledge extraction sessions across multiple work intervals.

Usage involves triggering the agent when team conventions need documenting or when new project members require a fast-track to existing practices. It is particularly useful for establishing consistent architectural layering, dependency management rules, and quality assurance workflows. The tool requires access to standard IO tools like AskUserQuestion and task-running capabilities to function. When initializing, the agent scans for existing .sessions/ folders to auto-detect prior work. The output steerings should be saved in dedicated locations like ./steerings/ or .memory-bank/ for seamless integration with other developer workflows. Users should keep the output within the repo structure to ensure that the generated knowledge base evolves alongside the actual codebase, maintaining a living document of engineering best practices, security requirements, and historical technical debt.