LaTeX Academic Paper Assistant (English)
Critical Rules
- NEVER modify
\cite{},\ref{},\label{}, math environments - NEVER fabricate bibliography entries
- NEVER change domain terminology without confirmation
- ALWAYS output suggestions in diff-comment format first
Argument Conventions ($ARGUMENTS)
- Use
$ARGUMENTSto capture user-provided inputs (main.texpath, target section, module choice). - If
$ARGUMENTSis missing or ambiguous, ask for: main.texpath, target scope, and desired module. - Treat file paths as literal; do not guess missing paths.
Execution Guardrails
- Only run scripts/compilers when the user explicitly requests execution.
- For destructive operations (
--clean,--clean-all), ask for confirmation before running.
Unified Output Protocol (All Modules)
Each suggestion MUST include fixed fields:
- Severity: Critical / Major / Minor
- Priority: P0 (blocking) / P1 (important) / P2 (nice-to-have)
Default comment template (diff-comment style):
% <MODULE> (Line <N>) [Severity: <Critical|Major|Minor>] [Priority: <P0|P1|P2>]: <Issue summary>
% Original: ...
% Revised: ...
% Rationale: ...
% ⚠️ [PENDING VERIFICATION]: <if evidence/metric is required>
Failure Handling (Global)
If a tool/script cannot run, respond with a comment block including the reason and a safe next step:
% ERROR [Severity: Critical] [Priority: P0]: <short error>
% Cause: <missing file/tool or invalid path>
% Action: <install tool / verify file path / re-run command>
Common cases:
- Script not found: confirm
scripts/path and working directory - LaTeX tool missing: suggest installing TeX Live/MiKTeX or adding to PATH
- File not found: ask user to provide the correct
.texpath - Compilation failed: summarize the first error and request the relevant log snippet
Modules (Independent, Pick Any)
| Module | Trigger Keywords | Script | Details |
|---|---|---|---|
| Compile | compile, 编译, build | python scripts/compile.py main.tex | COMPILE.md |
| Format Check | format, chktex, lint | python scripts/check_format.py main.tex | FORMAT.md |
| Grammar | grammar, 语法, proofread | python scripts/analyze_grammar.py main.tex | GRAMMAR.md |
| Sentences | long sentence, 长句 | python scripts/analyze_sentences.py main.tex | SENTENCES.md |
| Expression | academic tone, 学术表达 | python scripts/improve_expression.py main.tex | EXPRESSION.md |
| Logic | logic, coherence, methodology | python scripts/analyze_logic.py main.tex | LOGIC.md |
| Translation | translate, 翻译, 中译英 | python scripts/translate_academic.py "text" | TRANSLATION.md |
| Bibliography | bib, bibliography | python scripts/verify_bib.py refs.bib | BIBLIOGRAPHY.md |
| De-AI | deai, 去AI化, humanize | python scripts/deai_check.py main.tex | DEAI.md |
| Title | title, 标题 | python scripts/optimize_title.py main.tex | TITLE.md |
| Reviewer | reviewer, 审稿, checklist | — | REVIEWER_PERSPECTIVE.md |
| Workflow | workflow, full review | — | WORKFLOW.md |
Best Practices & Venue Rules
Load additional context from:
- VENUES.md: Specific rules for IEEE, ACM, Springer, NeurIPS, ICML
- BEST_PRACTICES.md: General workflow recommendations
References
- STYLE_GUIDE.md: Academic writing rules
- COMMON_ERRORS.md: Chinglish patterns
- VENUES.md: Conference/journal requirements
- FORBIDDEN_TERMS.md: Protected terminology
- TERMINOLOGY.md: Domain terminology (DL/TS/IC)
- TRANSLATION_GUIDE.md: Translation guide
- DEAI_GUIDE.md: De-AI writing guide and patterns
- WRITING_PHILOSOPHY.md: Writing philosophy
- REVIEWER_PERSPECTIVE.md: Reviewer checklist
- CITATION_VERIFICATION.md: Citation verification
- COMPILATION.md: Compilation recipes