paper-explainerlisted

# Paper Explainer ## Shared narrative rules Read and follow [the shared narrative and evidence rules](../shared/narrative-and-evidence.md). This skill adds paper-specific evidence and review rules but does not duplicate the shared wording policy. ## Overview Create a self-contained paper explanation that lets a reader understand the paper without repeatedly returning to the original. The explanation must answer not only “what modules exist”, but also “what each module receives, what it filters or changes, what state it updates, what it outputs, and why the next stage needs that output”. The target reader is an **average reader in the paper's relevant subfield**. Before drafting, make that audience concrete with two short internal lists: `assumed_known` for concepts the reader may be expected to know, and `explain_in_draft` for concepts that must be introduced. Never place paper-specific names, symbols, modules, prompt fields, state transitions, or implementation choices in `assumed_known`. The Main agent owns source collection, outlining, writing, revision, and final decisions. Two read-only subagents provide independent checks: - **Reviewer** verifies factual accuracy and coverage against all in-scope evidence. Load [references/reviewer-agent.md](references/reviewer-agent.md). - **Reader** simulates a human reader and checks comprehension, information density, and hidden assumptions using only the draft. Load [references/reader-agent.md](references/reader-agent.md).