core

renorm:core — Claim Discipline

Two contracts apply whenever this skill fires: tagging and grounding. Both are non-negotiable.

Tagging contract

Every substantive claim Claude makes must be tagged with one of:

• Derived — follows from explicit equations or derivation visible in this conversation.
• Assumed — explicit assumption being made for the discussion. State the assumption.
• Phenomenological — fitted or empirical input, not derived.
• Exploratory — speculative; not grounded in derivation or literature.
• Training-knowledge — Claude's recall of training material; not verified, possibly stale.

Untagged substantive claims are not allowed. If a claim doesn't fit one of these tags, the claim is not ready.

The tags apply to claims about physics content, not conversational scaffolding. "Let me think about this step by step" is not a claim; "the cuprate pseudogap is a Mott physics effect" is.

Grounding contract

Before making claims that depend on prior work (specific systems, mechanisms, recent results), present grounding first:

1. Default — training-knowledge. State understanding from training, tagged Training-knowledge. Identify:

   • What's known about this topic.
   • What's been tried.
   • What's adjacent in similar systems.

2. Escalation — only when user asks to "verify", "check", or "look up", or when the claim is obviously high-stakes:

   • Zotero MCP first. The user's curated library is the best first source — search it for relevant papers, fulltext, and the user's own annotations.
   • WebSearch / arXiv lookup only if Zotero comes up empty.

3. No external skill imports. Anything borrowed from third-party skills (e.g., yy/claude-scholar) gets copied into this plugin with attribution at port time, not wired as a runtime dependency.

Each claim points at its grounding (paper, equation, training-knowledge), or honestly says "this is my own connection, not in the literature."

What this skill does not do

• It does not fire on textbook recall ("explain BCS"), pure tooling questions, code debugging, or casual chat.
• It does not fire on algebraic steps inside an active derivation — those are governed by renorm:derivation (which references this skill's tagging vocabulary by name).
• It does not fire on paper-specific reading questions — those are governed by renorm:paper-analysis.

The wrappers reference this vocabulary by name; do not duplicate the contract in their SKILL.md files.

Activation via slash command

If the user explicitly invokes /renorm <claim>, apply both contracts to the supplied claim immediately. If /renorm is invoked without an argument, apply both contracts to the next exchange.