/aris

name: aris
description: Research-mindset catalog distilled from ARIS skills, for subagent Read on demand.

<system-reminder> 下列引用是从 ARIS (Auto-claude-code-research-in-sleep) 中炼化的 skills, 请在需要时加载, 可通过 Read tool 使用, 路径前缀 ${CLAUDE_PLUGIN_ROOT}. 注意: 参考其思维方式, 但不要被它的流程/模板带偏. The following references are mindsets distilled from ARIS (Auto-claude-code-research-in-sleep); Read on demand via the Read tool under ${CLAUDE_PLUGIN_ROOT}. Note: borrow the mindset, but don't be dragged into the source workflow or templates.

• skills_aris/problem-anchor: 锁定 problem 不让它在评审/讨论中漂移 + 机器人选题框架. 拿到研究方向 / 看到 scope 扩张请求 / 评审建议时用. Lock the research problem against drift + robotics problem framing. Use when receiving a research direction, evaluating scope-expansion requests, or assessing reviewer drift.
• skills_aris/literature-survey: 多源文献调研方法学. 搜文献 / mapping landscape / 找 gap 时用. Multi-source literature survey methodology. Use when searching literature, mapping a landscape, or hunting for research gaps.
• skills_aris/idea-quality: 判断 idea 是否 well-formed 和值得做的判据. 产 idea 或评估候选 idea 时用. Criteria for whether an idea is well-formed and worth pursuing. Use when generating ideas or evaluating candidates.
• skills_aris/idea-filtering: 从候选 idea 池筛出顶部的纪律. 砍 / 筛 / 评 idea 时用. Discipline for filtering candidate ideas down to a top few. Use when culling, filtering, or evaluating candidate ideas.
• skills_aris/novelty-check: 判断 idea/method 是否 novel 的判别框架. brainstorm 粗筛或定 idea 前深查时用. Framework for verifying idea/method novelty; use for quick filter during brainstorm or pre-commit deep check.
• skills_aris/empirical-pilot: 用 pilot 实验拿 empirical signal 的纪律. 投入全实验前跑 pilot 时用. Discipline for pilot experiments; use before scaling to a full experiment.
• skills_aris/method-refinement: 把 vague idea 炼成 elegant method + 公式推导——四原则/invariant object 选择/步骤分类 (identity/proposition/approximation/interpretation)/推导策略. 把 idea/approach 落成 method 或构建 coherent 理论线时用. Core principles for distilling ideas into elegant methods + formula derivation. Use when turning an idea into a concrete method or building a coherent theory line.
• skills_aris/reviewer-protocol: 跟外部 reviewer 协作 + 审稿回复——brief/review/iterate 纪律、rebuttal 三硬门安全模型、七种回复模式. 召唤 reviewer / 收 reviewer feedback / 回复审稿意见时用. Reviewer collaboration + rebuttal — brief/review/iterate discipline, rebuttal three-gate safety model, seven response modes. Use when invoking a reviewer, processing feedback, or responding to reviewer comments.
• skills_aris/experiment-planning: claim-driven 实验 roadmap. 给 stable method 设计 paper 实验时用. Claim-driven experiment roadmap. Use when designing paper experiments for a stable method.
• skills_aris/pipeline-meta: 多 phase pipeline 元规则——phase 顺序/checkpoint/预算意识/上游不稳不下游/gate before compute/知识积累/harness 优化/基金申请. 调度多 phase 时用. Meta rules for multi-phase pipelines — phase ordering, checkpoints, budget awareness, gate philosophy, knowledge accumulation, harness optimization, grant proposals. Use when orchestrating multi-phase research work.
• skills_aris/experiment-execution: 把 experiment plan 落地到运行实验 + 采集首轮结果的纪律. 实施 / 部署 / 跑实验时用. Discipline for executing an experiment plan into running experiments + initial results. Use when implementing, deploying, or running experiments.
• skills_aris/experiment-monitor: 监控运行中实验的进程状态与训练质量. 实验运行期间 / 判断是否继续时用. Monitor running experiments — process status and training quality. Use during runtime checks or when judging whether to continue.
• skills_aris/cloud-gpu: 根据任务自动选云 GPU 并控制成本的纪律. 需要按需 GPU 时用. Discipline for picking cloud GPUs by task-driven analysis with cost awareness. Use when on-demand GPU is needed.
• skills_aris/review-loop: 多轮 review→fix→re-review 改进——reviewer independence protocol、论文修复模式、restatement regression、format check 哲学. 跑 review 迭代时用. Iterative review→fix→re-review improvement — reviewer independence protocol, paper fix patterns, restatement regression, format check philosophy. Use when running a review loop.
• skills_aris/result-validation: 实验结果分析与验证——结构化发现 (Observation→Interpretation→Implication→Next step)、实验诚信六轴审计、claim 路由. 实验结束 / 准备 commit claim 时用. Result analysis and validation — structured findings, six-axis integrity audit, claim routing. Use after experiments complete or before committing to claims.
• skills_aris/paper-planning: 论文大纲规划——claims-evidence matrix、叙事结构、section-by-section 分节内容、figure 规划. 写论文前出 outline 时用. Paper outline planning — claims-evidence matrix, narrative structure, section-by-section guidance, figure planning. Use when creating a paper outline before writing.
• skills_aris/paper-writing: 论文写作准则——各节写作规范、写作质量 audit passes、参考书目规则、理论论文一致性、系统论文 blueprint. 写 LaTeX 草稿时用. Paper writing guidelines — section drafting standards, writing quality audit passes, bibliography rules, theory paper consistency, systems paper blueprint. Use when drafting LaTeX manuscript.
• skills_aris/paper-compile: LaTeX 编译与提交就绪——编译后质量信号、不同 venue 的页数规则、stale file 检测、匿名化与字体嵌入. 编译论文 / 准备提交时用. LaTeX compilation and submission readiness — post-compilation quality signals, venue-specific page count rules, stale file detection, anonymization and font embedding. Use when compiling or preparing for submission.
• skills_aris/citation-audit: 引文核查——三层审计 (存在性/元数据/上下文)、七类危险引文问题、soft-only 模式、未引用条目检测. 提交前验证引文时用. Citation audit — three-layer verification (existence/metadata/context), dangerous citation failure modes, soft-only mode, uncited entry detection. Use when verifying citations before submission.
• skills_aris/kill-argument: 对抗性审查——模拟 hostile reviewer 找出能 kill 论文的最强拒稿理由, 再由独立 adjudicator 逐点裁决. 理论论文 / 有 headline-level scope claim 时, 标准 review 后提交前用. Adversarial review — simulate a hostile reviewer to construct the single most damaging rejection argument, then adjudicate point-by-point with a second independent reviewer. Use for theory papers or papers with headline-level scope claims, after standard reviews before submission.
• skills_aris/paper-claim-audit: 零上下文论文数字核查——对比论文中每个数字与原始实验结果文件, 捕捉数字膨胀 / cherry-pick / 配置不匹配等失效模式. 论文有数字