What problem does it solve? When you ask an AI to implement a small feature directly, it often writes code blind—ignoring lessons already learned, decisions already made, and architectural boundaries already defined in the project. This Skill fixes that by requiring a quick search of the CodeStable knowledge base before any code is written, without imposing heavyweight design documents or checklists. ## Core Features & Use Cases - Pre-coding knowledge lookup: Searches codestable/compound for learnings, tricks, and decisions, plus codestable/architecture for subsystem boundaries, using search-yaml.py queries. - Lightweight coding guardrails: Enforces minimal-change principles—write only what was asked, place new logic in new files, avoid patch-style special-case branches, and stop at reflection signals like 300+ line files or 4+ function parameters. - Escalation path: Detects when a task outgrows fast-forward mode (3+ subsystems affected, new terminology, boundary changes) and hands off to the full cs-feat-design flow. - Use Case: You say "fast forward, add a filter to the serial monitor." The AI first searches past learnings and decisions about the monitor module, then implements the filter following established conventions—no design doc, no confirmation meetings. ## Quick Start Tell the AI "use fast forward mode to add a small feature" and describe the small change you want implemented directly.
