journal-cover-letter-skill

Journal Cover Letter Skill v3.2

Turn verified manuscript facts into a clear editorial decision aid. Default to English unless the user requests another language. Prioritize biomedical and life-science submissions while remaining useful across disciplines.

Core architecture: evidence anchor -> editorial meaning

Every persuasive claim must contain both parts:

• empirical_anchor: the concrete, manuscript-traceable observation selected for the editor;
• editorial_meaning: what that observation changes in understanding, interpretation, coordination, practice, or the next research decision.

For manuscripts with a distinctive empirical taxonomy, trend set, or named result pattern, also preserve an authorial_empirical_fingerprint: the minimum manuscript-native detail that lets an editor distinguish this paper from another paper on the same topic.

A letter fails when it offers facts without meaning, meaning without a traceable factual anchor, or a polished abstraction that erases the manuscript's empirical fingerprint. This is the central v3.1 rule.

Core rules

• Treat the manuscript and author-confirmed materials as the factual source of truth. Never invent titles, results, registrations, declarations, author details, editor names, or journal requirements.
• Separate facts into verified, conflict, missing, and not_applicable. Separate factual claims from interpretation.
• Use the strongest wording the evidence supports. Accuracy should sharpen the pitch, not make it timid.
• Separate the journal's official submission label from the manuscript's intellectual route. They may differ.
• Treat previous letters as user-controlled. Reuse or analyze them only within explicit permission.
• Treat an expert-authored letter as evidence of selection and editorial judgment, not as a gold standard or factual authority.
• When testing a skill against a human benchmark, freeze a blind baseline before revealing the benchmark, compare editorial effects rather than wording, and modify transferable rules rather than patching a single draft.
• Verify current journal information from official sources after the factual foundation is stable.
• Use bounded revision loops. Reaching a loop limit is not success.
• Use scripts for deterministic extraction, validation, auditing, and DOCX generation. Keep scientific meaning and editorial judgment in model reasoning.