$ show rsmdt/the-startup
The Agentic Startup repository is a spec-driven development (SDD) framework for Claude Code that enables multi-agent AI collaboration for software development. It provides a structured workflow: specify → validate → implement → test → review → document, with optional skills for analysis, refactoring, debugging, and constitution enforcement. The framework uses three implementation tiers (Direct, Incremental, Factory) selected by an automatic complexity classifier, and enforces quality gates through validation checks, drift detection, and constitution-based governance rules (L1/L2/L3 levels). Key features include progressive disclosure for context efficiency, information barriers to prevent bias, and automatic documentation generation. The system is designed as a plugin-based architecture with 'start' (core workflow) and 'team' (specialized agents) components, storing specifications in .start/specs/ and generating documentation in docs/.
The LLM identified 4 files worth highlighting for reuse:
Manages auto-incrementing spec IDs, directory structure (.start/specs/[NNN]-[name]/), and decision logging in spec README.md. This self-contained utility handles spec lifecycle management and could be extracted for use in other spec-driven systems with minimal modification. Demonstrates non-obvious skill in balancing automation with human oversight through decision tracking.
Provides real-time feedback on context usage, session cost, git branch, and active spec ID with color-coded thresholds. This self-contained monitoring script is configurable via TOML and could be extracted for use in other CLI tools. Shows craftsmanship in balancing informative output with minimal performance overhead.
Defines a 6-rung debugging escalation framework from path tracing to interactive inspection. This reusable methodology provides a structured approach to root cause analysis that could be applied to any software system. Demonstrates non-obvious skill in creating a progressive disclosure system for debugging that matches effort to problem complexity.
Catalog of code smells with corresponding refactoring techniques and behavior preservation checklists. This self-contained reference provides actionable guidance for improving code quality that could be used independently of the framework. Shows craftsmanship in connecting theoretical concepts to practical refactoring steps with clear examples.