mdmanurung/lucid-writer

Turn one lasting writing correction into one proposed lucid-writer rule, or propose rules from pasted bench ratings. Use when the user says "remember this", "always write X", "never write Y", "from now on", "save this as a writing rule", runs /lucid-writer:learn, or accepts the offer to save a correction. With --from-ratings, read pasted calibration and rewrite ratings with notes and draft proposals. Claude Code only; the rules are injected by plugin hooks.

This skill should be used when the user asks to draft, revise, tighten, restructure, or review a research article, scientific or technical report, abstract, manuscript section, reviewer response, or submission material; asks to turn figures, tables, analyses, or notes into scholarly prose; asks to check claim–evidence alignment, causal language, citations, reporting completeness, coherence, or generic AI/LLM-sounding academic prose; or asks to judge whether evidence supports a descriptive, associative, predictive, causal, or mechanistic claim, or whether a statement is an observation, literature fact, inference, or hypothesis. Do not use for casual, creative, marketing, or personal writing.

List and manage learned lucid-writer writing rules. Approve, reject or retire one rule, or trust a project's .lucid-writer/rules.json. Use when the user runs /lucid-writer:rules, asks "what writing rules have you learned", "show my rules", "approve r-0003", "drop that rule", or wants to trust a project's rules. Claude Code only.