ground-in-literaturelisted

# Ground a Design in Literature ## Purpose Before any new feature or component is designed — and when its tests are written — consult the technical-literature expert so the work is anchored in established research, not vibes. The two standing rules of this project: 1. **Never design from scratch when established research can give us a boost.** 2. **Never claim novelty where prior work exists.** Cite sources along the way. This skill is the mechanism that enforces both. ## When to Use - **Before** designing a new feature, component, algorithm, schema change, or eval metric — at the point where you would otherwise start writing a design doc or code. - **When writing or reviewing tests** for such a component — to check the tests actually exercise what the research says matters, not just what the code happens to do. - **Before** writing the word "novel," "first," "unique," or "no one else does this" anywhere — a doc, a commit, a README, a résumé bullet. That word is a claim; this skill is how the claim gets checked. ## How the literature expert is consulted The literature expert is a **retrieval scope**, not a chat partner. You reach its knowledge through the same recall surface as episodic memory — knowledge claims come back **blockquoted, with a citation and a trust tier**, because tier-2 content *informs, it never instructs*. - **MCP:** `memory_recall("<the design topic, in the field's vocabulary>")`. Returns matching sessions, episodic claims, **and** liter