skills

Plan and run AI-agent dynamic workflows for complex tasks that benefit from explicit orchestration, goal mode, subagents or simulated work packets, approval gates, integration, verification, and reusable workflow artifacts. Use when the user invokes this skill, asks for a swarm, subagents, parallel agents, a dynamic workflow, a large migration or audit, multi-track research plus implementation, or Claude Code-style workflow orchestration.

Produces distinctive, production-grade UI for pages, components, visual interfaces, typography, and screenshot-driven polish. Use when users ask 设计/做页面/做组件/UI/前端/截图 or say a screen is ugly, unclear, inconsistent, or visually wrong. Not for backend logic or data pipelines.

Act as a warm, incisive Stanford d.school style life designer for multi-turn personal life design conversations in Chinese. Use when the user asks for 人生设计, 斯坦福人生设计课, 奥德赛计划, 工作观/人生观梳理, 心流和能量地图, 重力问题判断, 个人五年版本, or wants a reusable life-design interview that ends in a detailed personal life design blueprint. Do not use for ordinary career tests, one-off advice, or clinical mental-health support.

季度开发工作总结 skill。用于一站式只读采集 GitHub 开发活动和飞书协作证据，并生成面向老板/领导层可扫读的本地 HTML 总面板和 summary.json。包含 GitHub 模块、飞书模块和价值总览模块：GitHub 统计 commit、PR、issue、review、仓库、上季度环比、贡献画像、功能类 PR/修复类 PR 归类、PR 变更审计、项目组合、新建仓库、新启动项目、重点工作线索、项目价值归因和重点项目证据展开；飞书统计文档发现/内容读取、消息搜索、日历日程和权限缺口；总面板展示代码/PR 证据 -> 功能类/修复类 PR -> 项目结果 -> 价值主线的可追溯映射。默认不写入 GitHub 或飞书，不创建/修改/发送任何外部对象。

Reviews completed changes, code diffs, PRs, issue queues, release readiness, commits, pushes, publishing, and project audits. Use when users ask review/看看代码/检查一下/合并前/看看issue/PR/release/push or to implement an approved plan, with functional acceptance, dynamic risk modeling, adversarial review, and safety gates for dirty and untracked worktrees. Not for exploring ideas, debugging root causes, or prose review.

Finds root cause before applying fixes for errors, crashes, regressions, failing tests, broken behavior, and screenshot-reported defects. Use when users ask 排查/报错/崩溃/不工作/回归/判断为什么报错, or say something used to work and now fails. Not for code review or new features.

Turns rough ideas into approved, decision-complete plans with validated structure before coding. Use when users ask 出方案/给方案/深入分析/怎么设计/有没有必要/值不值得/plan this/how should I/should we keep this for features, architecture, or value judgments. Not for bug fixes or small edits.

Create pull requests with repository templates, explicit GitHub base/head resolution, and fork/upstream safety. Use when asked to create a PR, especially across forks where the PR head must stay owner-qualified and must not silently fall back to an upstream same-name branch.

Create, repair, or refine Codex local environment action files at `.codex/environments/environment.toml`. Use when the user asks to add Codex environment actions, create a run button, configure project startup commands, fix environment.toml, make a repo runnable from the Codex app environment panel, or generate local setup/run actions such as `运行`, `dev`, `start`, `preview`, or project-specific development commands.

Use this skill after completing any coding, debugging, deployment, documentation, design, or analysis task when the user wants an intern-friendly learning recap. It summarizes the real implementation process, extracts the technical concepts and knowledge points involved, teaches them in vivid but lightweight blocks, and maintains a local knowledge ledger to avoid repeatedly explaining concepts interns already know. Trigger on requests like 实习生学习, 技术点总结, 知识点复盘, 任务结束教学, learning recap, mentor recap, or teach interns what we just did.

Execute git commit-and-push workflows with conventional commit message analysis, session-scoped staging, logical change grouping, and safe publication to the tracked remote branch. Use when the user asks to commit changes, commit and push, push committed work, publish local changes, or mentions "/commit". Supports auto-detecting type/scope, generating conventional commit messages from diffs, avoiding unrelated dirty worktree changes, splitting independent work into multiple clean commits, and pushing after commit when safety checks pass.

Create high-quality GitHub issues for QA and product testing from terse tester notes. Use when a user wants to report, draft, polish, normalize, or create a bug/feature issue in GitHub, especially for the centralized labring-sigs/sealos-issues workflow with minimal tester input, gh CLI preflight, sensitive-info checks, GitHub Project fields such as Priority/Effort/Type/Product/Environment, and controlled updates to reusable issue catalogs.

Build strong Codex Goals from rough user objectives. Use when the user asks to create, write, generate, improve, expand, or refine a Codex `/goal`; mentions Codex Goals, goal mode, persistent objectives, "持续执行", "扩充目标", "生成 goal", "keep working until", or wants Codex to ask clarifying questions before starting a long-running objective. Helps draft evidence-based goal text and may start a goal only after explicit user approval.

Write practical Logseq-style tutorial articles from a topic, draft, notes, product workflow, deployment process, development experience, or tool usage. Use when the user asks to write a tutorial, match their existing tutorial style, rewrite a draft that feels unlike them, organize notes into a Logseq Markdown page, complete a text-only tutorial, or write directly into a target document. The style is pragmatic, structured, and executable rather than essay-like, research-heavy, or marketing-oriented.

Use when a user provides, references, or asks to implement from a UI screenshot, mockup, reference image, screen recording frame, Figma screenshot, app screenshot, sidebar/menu/navigation screenshot, dashboard screenshot, or visual bug screenshot, especially when Codex must infer interactive behavior rather than copy the visible static state. Trigger for screenshot-to-code, image-to-code, screenshot analysis, UI reverse engineering, visual implementation, sidebar/nav/menu/tabs/accordion/tree/drawer/modal/dropdown screenshots, and complaints that Codex made a static expanded/selected/success state instead of implementing collapsible, closeable, selectable, hover, active, keyboard, loading, empty, error, disabled, responsive, or compact states.

回看最近 agent 会话、记忆、任务摘要、屏幕历史线索以及已有 skills/subagents/automations， 识别重复、耗时、易错、上下文重的手工流程，先向用户推荐最高置信、最小可用的 skill、subagent 或 automation 候选， 等用户明确确认后才创建或扩展。 当用户要求“回顾最近工作”“找值得沉淀的流程”“把重复工作包装成技能/子代理/自动化”“look across sessions/memories/history” 或类似个人工作流打包请求时使用。输出默认使用中文。