菜单
SkillsMP 已收集 Ocean326/Agents 中的 47 个 Skill。打开任一 Skill 可查看来源和详情。
已展示 40 / 47 个已收集 Skill。
Design single-agent or multi-agent workflows, orchestration loops, handoffs, evaluator-optimizer cycles, and wake-up or re-entry patterns for agentic systems. Use when Codex needs to decide whether work should stay as a prompt, instruction, skill, single…
原文语言：英语
面向架构类工作的总入口路由器。用于在广义架构总控、ADR/治理、现有代码库重构、规格驱动设计、DDD/clean architecture、agent systems architecture 与架构文档评审之间选择最合适的入口；当用户知道“这是架构问题”但不知道具体 lane 时优先使用。
原文语言：英语
Senior architect entrypoint that routes software architecture, 系统设计, 技术方案, 架构评审, ADR, 代码库重构, requirements-driven design, DDD/clean architecture, agent system design, and architecture-document review to the right Architecture Workbench skill. Use when Codex…
原文语言：英语
Attach a PersonalBrain repository to the current task using a minimal, upgrade-friendly context rooted at the PersonalBrain path. Use when Codex should treat a PersonalBrain vault as shared durable memory for routing, retrieval, external-project awareness, or…
原文语言：英语
Use when the user wants to shape an ambiguous feature, product idea, workflow, or UX direction before implementation, especially when requirements, trade-offs, or success criteria are still unclear. Do not use for straightforward bug fixes, direct…
原文语言：英语
面向前端、设计工具链、部署、GitHub 工作流与视觉资产生成的总入口路由器。用于在 visually strong UI、GitHub 操作、Vercel 部署、Figma/Canva 工作流、以及 bitmap image generation/editing 之间选择最合适的技能；当用户只知道“要把东西做出来并上线/出图/进 GitHub”时优先使用。
原文语言：英语
Sync, validate, and operate Butler-style chat connectors on this machine, including a local Feishu bot config, the Weixin bridge and QR login flow, and the newer official Lark/Feishu CLI and MCP routes. Use when Codex needs to migrate Butler chat capability…
原文语言：英语
Build robust local command-line tools and lightweight interactive operator surfaces. Use when Codex needs to turn a workflow into a reusable CLI, add subcommands or flags, wrap scripts with safer argument handling, create a small terminal UI or guided prompt…
原文语言：英语
Review historical Codex sessions in `~/.codex/session_index.jsonl`, `~/.codex/history.jsonl`, and `~/.codex/archived_sessions/rollout-*.jsonl` to discover recurring workflow, prompt, and skill issues across past conversations. Use when Codex needs to turn…
原文语言：英语
Compile fragmented local and external materials into report-grade deliverables with local-first evidence mapping, adaptive web discovery, and explicit separation of facts, judgments, inferences, decisions, and next actions. Use when Codex needs to turn…
原文语言：英语
Coordinate recoverable remote SSH compute from any local project by using the central Conductor runtime at <PERSONALBRAIN_ROOT>/.local/conductor, selecting a verified SSH route for either a workstation or a shared compute server, dispatching direct remote…
原文语言：英语
Use DeepXiv as an agent-first paper search and progressive reading lane for arXiv, PMC, bioRxiv, and medRxiv friendly literature. Best for fast paper discovery, `brief -> head -> section -> full` reading, and identifier recovery before a local library intake…
原文语言：英语
High-autonomy delivery orchestrator that takes work from an idea, bug report, refactor request, or partially defined project to verified delivery by sweeping context, choosing the right execution path, coordinating product/design/implementation/testing…
原文语言：英语
面向交付与产品工作的总入口路由器。用于在端到端交付、产品发现、方案生成、需求定义、PRD/tech spec、方案批判、测试用例与最终验证之间选择最合适的技能；当用户要“把事情做成”但尚未明确是 discovery、proposal、planning、implementation 还是 verification 时优先使用。
原文语言：英语
面向执行编排的总入口路由器。用于判断当前任务是否应该保持在本线程本地执行，还是需要远端 SSH 计算、显式并行代理、或按计划拆给实现子代理；当用户关心“怎么执行这项工作”而不是“这项工作属于哪个内容领域”时优先使用。
原文语言：英语
Extract memory, decisions, key outputs, and pending work from one or more Codex or Cursor chat session IDs, then continue execution in the current thread with a concrete next-step plan. Use when user says resume/continue session, hand over by chat id, recover…
原文语言：英语
Apply an evidence-shaped Huang Chao / HKUDS perspective to agent systems, RAG, coding or research assistants, and open-source product strategy. Use when Codex should critique or shape an idea through workflow leverage, agent-native interfaces, simple…
原文语言：英语
面向知识接入、上下文挂接、官方文档检索与 prompt 结构化的总入口路由器。用于在 PersonalBrain 挂接与检索、OpenAI 官方文档查询、以及 prompt 优化之间选择最合适的技能；当用户知道需要“先补上下文/先查资料/先把 prompt 变稳”但不知道具体技能名时优先使用。
原文语言：英语
End-to-end paper discovery, identifier recovery, accessible PDF download, and local literature library bootstrapping. Use when Codex needs to turn BibTeX, EndNote `.enw`, RIS, DOI or arXiv lists, title batches, or a vague "帮我把这些文献找齐并落到本地" request into local…
原文语言：英语
Build, export, and validate local Docker packaging artifacts for arbitrary repositories on this Mac. Use when Codex needs to ensure Docker or Colima is running, discover project-specific Docker build or bundle scripts, build an image from a Dockerfile, export…
原文语言：英语
Apply Ocean's evidence-shaped operating constraints and taste to planning, critique, research shaping, governance, and delivery. Use when work should stay exact in scope, evidence-first, local-artifact-first, stage-ordered, traceable, and conservatively…
原文语言：英语
Bulk data enrichment. Adds web-sourced fields (CEO names, funding, contact info) to lists of companies, people, or products. Use for enriching CSV files or inline data. Supports multi-turn: pass --previous-interaction-id from a prior research or enrichment to…
原文语言：英语
URL content extraction. Use for fetching any URL - webpages, articles, PDFs, JavaScript-heavy sites. Token-efficient: runs in forked context. Prefer over built-in WebFetch.
原文语言：英语
DEFAULT for all research and web queries. Use for any lookup, research, investigation, or question needing current info. Fast and cost-effective. Only use parallel-deep-research if user explicitly requests 'deep' or 'exhaustive' research.
原文语言：英语
Distill a real, historical, fictional, or privately known person from mixed source channels into a reusable skill. Combines multi-channel evidence intake with mindset, task-signature, taste-rubric, and persona extraction, then chooses the right output shape:…
原文语言：英语
Refine an existing proposal, draft, or partially formed plan through bounded multi-round critique, targeted improvement, and re-verification. Use when a coherent proposal already exists and needs critique rather than blank-page ideation or multi-option…
原文语言：英语
Generate 2-5 structured candidate proposal briefs from a vague goal, problem, or opportunity, then synthesize them into a critique-ready recommendation packet. Use when the user needs blank-page ideation to become multiple complete proposals across research,…
原文语言：英语
为 AI 研究规划考虑算力约束的实验方案，包括数据集、划分、基线、指标、消融和成功标准。当 Codex 需要把一个 idea 变成可执行的深度学习或序列表征学习实验计划时使用。
面向 AI 研究流程的总入口路由器，覆盖读论文、想法澄清、实验设计、结果分析和论文写作。当下一步不明确、任务跨越多个阶段，或 Codex 需要协调完整的深度学习与序列表征学习研究链路时使用。
将模糊的 AI、深度学习和序列表征学习想法整理成更清晰的目标、假设、可检验命题、对比维度和最小可行实验。当用户只有粗糙 idea、半成型方法或零散笔记，需要把它们变成研究方向时使用。
评估 AI、深度学习和序列表征学习 idea 或 draft 的新颖性与竞争定位。当 Codex 需要对比 prior work、压力测试贡献主张、提前暴露 reviewer 质疑，或判断一个想法是否值得继续做实验时使用。
将研究笔记和实验结果转成论文段落、摘要、相关工作定位、贡献表述、修改计划或 rebuttal，适用于 AI、深度学习和序列表征学习工作。当 Codex 需要起草、重构或修改论文时使用。
深度阅读 AI、深度学习和序列表示学习论文，结构化提取问题定义、核心主张、模型结构、训练目标、数据集、基线、消融和失败模式。当 Codex 需要读论文、比较多篇工作，或在立题、实验、写作前整理研究笔记时使用。
面向 AI 研究工作的总入口路由器。用于在查文献与建文献库、读论文、idea 澄清、新颖性压测、实验设计、结果分析、论文写作、slides/poster 与网页材料抓取之间选择最合适的下一步；当研究阶段不清楚、材料混杂、或用户只知道“想推进研究”时优先使用。
原文语言：英语
基于 AI 和深度学习材料构建 talk deck、读书汇报、组会更新、poster 和研究简报。当 Codex 需要整理报告叙事、图表顺序、视觉层级或口头表达结构时使用。
消化 AI 和深度学习论文中的 appendix、补充材料、长表格、实现细节和隐藏实验设置。当正文写得不够清楚，或 Codex 需要复现关键细节、消融设置、面向审稿人的补充证据时使用。
分析 AI、深度学习和序列表征学习任务中的训练过程、消融结果、评测表格和失败模式。当 Codex 需要解释模型行为、排序下一步实验，或把 runs 总结成清晰结论时使用。
Full-featured Agent Skills management: Search 35+ skills, install locally, star favorites, update from sources. Use when looking for skills, installing new skills, or managing your skill collection.
原文语言：英语
Asynchronously close the loop on skills used in the current thread by routing to PersonalBrain's canonical `skill_skill_usage_close_loop` workflow, spawning one bounded background subagent, and appending per-skill evidence rows for later aggregate analysis.…
原文语言：英语
面向 skill 体系运维的总入口路由器。用于在技能发现、安装、来源材料抽取成 skill、创建/更新、插件脚手架、以及 skill 使用闭环复盘之间选择最合适的技能；当用户要扩技能体系、整理 skills、决定该扩现有 skill 还是新建 skill，或把人/书/理论/博客蒸馏成 skill 时优先使用。
原文语言：英语