research_copilot

Final 6-dimension sanity check (logic/citation/reproducibility/novelty/venue/de-AI). Use before submission for comprehensive audit.

Pre-submission optimization loop (review → fix → review). Use before final submission to close all P0 gaps.

Polishes paper text and removes AI patterns. Use after writing stage to refine language and ensure venue compliance.

Designs and launches experiments. Handles long-running tasks via Monitor. Use for experiment tasks.

Focused literature search skill. Searches papers, locks baselines, builds related-work map. Use for standalone literature tasks.

Orchestrate a complete research workflow from literature review to camera-ready paper. Calls `rc` CLI commands to create tasks, dispatches agents for execution, verifies quality gates. Follows Trellis philosophy: Task-first, Action-before-asking, Single source of truth.

Use when setting up the plugin for the first time, installing dependencies, configuring MCP servers, or when the user says '初始化', 'init', 'setup', '装环境', '配置', 'install', 'configure', 'first time', '首次使用'. Handles both dependency marketplace installation and MCP server setup.

Pre-plan Socratic clarification. Use BEFORE drafting any plan (ideation direction, experiment Goal anchor, rebuttal strategy, pipeline routing) to expose hidden assumptions, lock topology, and converge on scope. Walks the decision tree one branch at a time with a recommended answer per question and an explicit ambiguity score, stopping only when the score crosses the threshold. Triggers on: '帮我厘清需求', '我有个想法不太明确', '在写计划前先聊清楚', '深度访谈', 'clarify before planning', 'deep interview', 'requirements interview', 'scope before plan'.

Research pipeline workflow enforcement. Use when coordinating any research stage (literature/ideation/experiment/writing/polishing/review/rebuttal). Provides mandatory checklist and state machine rules.

Workflow for Research Copilot plugin development. Use when creating, updating, reviewing, or debugging `.agent.md`, `SKILL.md`, agent orchestration rules, worker dispatch protocols, pipeline ledger workflows, or plugin packaging docs. Triggers on: plugin-dev, agent-development, agent development, skill development, pipeline ledger, worker dispatch, agent orchestration.

Validation skill for de-AI quality checking. Use AFTER polishing to verify that AI patterns have been removed and text reads naturally. Triggers on: 'check de-AI quality', 'verify humanization', 'validate polish', 'de-AI checker', '检查去AI效果'. This is a validation-gate compatible wrapper around paper-deai.

Post-plan stress test. Use AFTER a plan is drafted (Goal anchor, ideation candidate, rebuttal strategy, pipeline template) to gap-check the plan against the project's existing documentation, terminology, and recent reviewer / handoff history in `.copilot/`. Sharpens fuzzy terms inline, cross-references the codebase / tex / logs, and offers ADRs only when a decision is hard to reverse. Never used to draft the plan itself. Triggers on: '校验计划', '对着文档拷问', '把计划放到文档里盘一遍', 'grill the plan', 'stress-test plan', 'audit plan against docs', 'check plan for gaps'.

Use whenever the user wants to search for academic papers, find recent research, look up conference publications, or explore literature on any AI / ML / CV topic. Routes through the `arxivsub-search` MCP server (arXiv + CVPR / ICCV / ICLR / ICML / NeurIPS / AAAI / MICCAI). Triggers on: "find papers on X", "what are the latest papers about Y", "search arXiv for Z", "any recent work on W", '论文检索', '最新研究'.

Use when repeated debugging or writing iterations fail, root cause is unclear, environment limits block progress, the user is still dissatisfied after multiple attempts, or the user says '疑难杂症', '卡住', '多轮迭代无解', '反复失败', '更强模型', '升级求助', 'stuck', 'escalate', 'stronger model'. Produces a hand-off summary suitable for a stronger model to pick up.

Use when the user wants a paper's abstract + method turned into a publication-ready architecture diagram, rendered as a single HTML file with inline SVG (and Python-generated SVG sub-figures for heatmaps / distributions / scatter / matrices). Triggers on: "架构图", "结构图", "method figure", "overview figure", "pipeline diagram", "draw methodology", "网页绘图". Enforces compactness numerics and a mandatory 10-round self-check loop. Do NOT use for line / bar / scatter data plots, posters, art, pure Mermaid sketches, or before the method is settled.

Use when the user asks to de-AI English LaTeX prose, humanize it, or remove the mechanical / chat-bot patterns. Triggers on: "去AI味", "de-AI", "remove AI patterns", "humanize", "去机械化". Rewrites at the same word count — does not add or delete content. Do NOT use for translation, expansion, or shortening — those have their own skills.

Use when the user asks to translate English LaTeX paper text into Chinese for comprehension. Triggers on: "翻译成中文", "English to Chinese", "英译中", "translate to Chinese". Outputs natural-language Chinese (not LaTeX). Strips citations / refs / formatting commands.

Use when the user asks to lightly expand English LaTeX text — add depth, surface implicit reasoning, increase logical connection. Triggers on: "扩写", "expand", "enrich", "elaborate". Does NOT pad with filler. Do NOT use for shortening (paper-shorten) or translation (paper-translate).

Use when the user has ML experiment results and needs them interpreted into LaTeX analysis paragraphs. Triggers on: "实验分析", "analyze experiments", "experiment analysis", "分析数据", "data analysis". Comparison & trend-driven; not a numerical recitation. Do NOT use for plotting (separate skill) or pure copy-editing.

Use when the user asks for a publication-ready English figure caption from a Chinese description. Triggers on: "图标题", "figure caption", "figure title", "图的标题". Outputs the caption text only — no `Figure 1:` prefix. Do NOT use for table captions (paper-table-caption).

Use when the user asks for a red-line consistency / logic check on near-final English LaTeX. High tolerance — only fatal issues are reported. Triggers on: "逻辑检查", "logic check", "校对", "proofread", "consistency check". Do NOT use for style polishing (paper-polish) or pre-submission audit (paper-sanity-check).

Use when the user asks to deep-polish an English LaTeX paper or a Chinese grant proposal — rigorous academic register, reduced AI feel, structured contribution framing. Triggers on: "polish", "润色", "deep rewrite", "publication quality", "基金润色". Works in both English (top-conference papers) and Chinese (NSFC-style proposals). Do NOT use for translation, expansion, or shortening — those have their own skills.

Use when the user wants a top-conference reviewer perspective on paper quality, with a severe-but-constructive tone. Triggers on: "审稿", "review", "reviewer", "peer review", "评审". Outputs a review report + strategic author advice. Do NOT use for pre-submission sanity check (paper-sanity-check) or polish (paper-polish).

Use when the user is preparing to submit a paper or has completed a major revision and needs a pre-submission factual / structural / logical audit. Triggers on: "sanity check", "查错", "基础检查", "check paper", "verify paper", "论文检查", "pre-submission check". Six-pass audit. Do NOT use for writing style (paper-polish) or substantive review (paper-review).

Use when the user asks to lightly compress English LaTeX text without losing information (target 5-15 words removed). Triggers on: "缩写", "缩减", "shorten", "compress", "reduce length". Does NOT delete content; only tightens syntax. Do NOT use for expansion (paper-expand) or translation (paper-translate).

Use when the user asks for a publication-ready English table caption from a Chinese description. Triggers on: "表标题", "table caption", "table title", "表的标题". Outputs the caption text only — no `Table 1:` prefix. Do NOT use for figure captions (paper-figure-caption).

Use when the user has a Chinese academic draft and needs it converted to publication-ready English LaTeX. Triggers on: "中译英", "Chinese to English", "translate", "翻译成英文". Combined translation + light polish. Do NOT use for English-to-Chinese (use paper-en2zh) or deep rewriting (use paper-polish).

Use when the user asks to '开始实验', '推进实验', 'run experiment', 'advance the experiment', 'turn an idea into experiment', or wants an idea JSON converted into concrete code edits and runs directly in Copilot. Copilot-native — Copilot writes the changes and reads the results in-session; the terminal only executes non-model commands. Do NOT use for plotting (scientist-plotting) or writeup (scientist-writeup).

Use when the user has a workshop / topic Markdown and wants it turned into an AI-Scientist-format ideas JSON, generated directly in Copilot. Triggers on: '生成 ideas', 'topic 变成想法', 'AI Scientist 出点子', 'generate ideas from topic'. Copilot-native — no workspace ideation script call.

Use when the user asks to '聚合作图', '补图表', '整理实验图', or wants experiment outputs converted into plots + plotting scripts directly in Copilot. Triggers on: 'aggregate plots', 'make figures from results'. Copilot-native — Copilot designs the figure and edits the script; the terminal only runs the Python plotting code. Do NOT use without existing experiment outputs.

Use when the user asks to '审一下这篇 PDF', '自动审稿', '给我 review', or wants a manuscript / PDF reviewed in Copilot with structured feedback. Triggers on: 'review this manuscript', 'auto-review'. Copilot-native — no workspace review script call.

Use when the user asks to '检查环境', '能不能跑 AI Scientist', 'runtime check', '初始化 AI Scientist 环境', or wants the ai-scientist MCP to validate Python / CUDA / LaTeX / poppler / runtime prerequisites. Routes through the `ai-scientist` MCP `validate_runtime`. Do NOT use as a substitute for actually running an experiment.

Use when the user asks to '开始写论文', '生成 PDF', '整理成论文', or wants LaTeX / Markdown drafted directly in Copilot from experiment artifacts. Triggers on: 'write the paper', 'generate PDF', 'compile to paper'. Copilot-native — no workspace writeup script call. Do NOT use for review (scientist-review) or plotting (scientist-plotting).

Reply-style controller. Use this skill on every turn to keep responses direct, dense, and free of filler. Triggers on: every conversational context — load this skill as the default reply style, in any language. 对话风格控制技能。