写作小能手

Use this skill to convert Chinese notes, rough English, experimental results, figure descriptions, or draft paragraphs into polished biomedical manuscript English.

Core Rules

1. Preserve the student's scientific meaning. Do not invent experiments, statistics, mechanisms, controls, sample sizes, or citations.
2. Use the bundled knowledge base under data/ for wording decisions. If a useful external word, phrase, or sentence structure is needed, ask permission and label it as external.
3. Use source examples only as style evidence. Never copy source sentences verbatim.
4. Keep claims calibrated: use cautious wording for inference, association, preliminary findings, and unsupported mechanisms.
5. Ask for the missing experimental subject, comparison group, intervention, endpoint, statistical result, or manuscript section only when a safe rewrite is not possible.

Writing Function Router

Map the task to one or more functions before drafting:

Knowledge Base Workflow

Prefer the unified query entry point before drafting:

python3 scripts/query_knowledge_base.py --function 研究空白 --query "remains unclear" --limit 5
python3 scripts/query_knowledge_base.py --function 机制解释 --query "T cell" --mode auto --limit 5
python3 scripts/query_knowledge_base.py --query "pulmonary drug delivery" --mode terms --limit 8

Use source-specific scripts when a narrow lookup is needed. In this extracted-under-10MB SkillHub package, classified_sentences_full.csv and lab_lexicon.csv are expanded subsets; replace them with full CSV exports for complete local corpus search.

• scripts/search_patterns.py: compact rule-level templates from data/pattern_library.csv.
• scripts/search_sentences.py: full sentence-pattern library from data/classified_sentences_full.csv.
• scripts/search_collocations.py: conventional biomedical collocations from data/collocation_index.csv.
• scripts/search_terms.py: terminology, verbs, transitions, adjectives, adverbs, and abbreviations from data/lab_lexicon.csv.

See references/data-sources.md only when updating the knowledge base, debugging search behavior, or checking canonical CSV names.

Drafting Workflow

1. Identify manuscript section and writing function.
2. Query the knowledge base for matching templates, collocations, and terminology.
3. Select patterns that match the user's actual content and evidence strength.
4. Draft in clear subject-verb-object English with explicit comparison groups.
5. Replace placeholders such as <factor>, <outcome>, and <mechanism> with the user's actual content.
6. Provide a brief pattern or wording note when the user is a student or asks for learning-oriented feedback.

Output Patterns

For sentence-level help, return:

1. Polished version: one recommended sentence.
2. Alternative: one or two variants when useful.
3. Pattern note: the sentence skeleton, collocation, or wording choice used.
4. Caution: only when the original claim is too strong or lacks required context.

For paragraph-level help, return:

1. A polished paragraph.
2. Sentence functions used in order.
3. Missing scientific details, only if they affect accuracy.

For figure legends:

1. Start with what the figure shows.
2. Name model, treatment, control, assay, endpoint, and statistics when provided.
3. Keep interpretation out unless the user asks for an interpretive legend.

For abstracts:

1. Background or gap.
2. Objective.
3. Key methods.
4. Main results.
5. Calibrated conclusion or significance.

Style Rules

• Prefer specific biomedical verbs: promoted, attenuated, enhanced, impaired, induced, mediated, regulated, was associated with.
• Use suggest, indicate, or support when evidence is indirect.
• Use demonstrate only when the experiment directly supports the claim.
• Avoid overclaiming words such as prove, fully elucidate, revolutionary, or first-ever.
• Avoid direct Chinese word order. Keep biomedical English concise, explicit, and logically sequenced.

Knowledge Base Maintenance

Use scripts/update_data.py only when replacing the bundled CSV knowledge base. It validates required headers, backs up existing canonical files, copies new CSVs into data/, and writes data/update_manifest.json.

python3 scripts/update_data.py --source-dir "/path/to/new/csv_exports"
python3 scripts/update_data.py --sentence-zip biomedical_sentence_candidate_files_all.zip --lexicon-zip vocabulary_frequency_gt5_by_category_csv.zip
python3 scripts/update_data.py --sentence-zip sentence_candidate_files_all_2026-05-21_pdf_set.zip --lexicon-xlsx vocabulary_frequency_gt5_by_category.xlsx --lexicon-sheet "总表_频率大于5"