What problem does it solve? Standing context for AI coding agents drifts: rules get duplicated across CLAUDE.md, skills, and output styles, stale lessons linger, and new instructions land in the wrong layer. This Skill routes every piece of context to its correct home and keeps the corpus free of conflicting duplicates. ## Core Features & Use Cases - Layer routing: Decides whether a lesson belongs in a global rules.d fragment, a repo CLAUDE.md, a skill, or nowhere, based on scope and audience. - Duplicate detection and consolidation: Classifies overlaps as undeclared duplicates, declared copies, or pointers, and picks the canonical home so each rule legislates in exactly one place. - Writing standards: Encodes the craft rules for skill descriptions, the global CLAUDE.md letter voice, and repo CLAUDE.md files, distilled from Anthropic's skill-creator and Claude context-engineering guidance. - Use Case: After Justin corrects the agent twice about the same behavior, use this Skill to save the generalized lesson as a rules.d fragment, then run the scoped sweep in resources/sweeps.md to confirm no twin exists elsewhere. ## Quick Start Ask the agent to remember a standing rule or audit the existing context files, and it will route the lesson to the right layer and sweep for duplicates before committing.
