my-ai-cli-toolkit

Create or review publication-ready academic figures that meet journal submission specs (IEEE / Elsevier / Nature, plus Springer and chinese-thesis presets) using matplotlib(+seaborn) or plotly; in industrytslib projects it drives that library's built-in visualization system. Use for 论文配图, 期刊图, 科研绘图, 中文学位论文配图, "publication-ready figure". Do NOT use to reproduce a specific paper's figure or style — that is paper-plot.

Audit and improve Codex AGENTS.md guidance files and companion code_map.md navigation maps. Use when the user asks to check, audit, optimize, or fix AGENTS.md; mentions nested AGENTS.md conflicts, code maps, stale commands, or scoped instructions; or says 优化 AGENTS.md, 审计 AGENTS.md, 更新 Codex 项目指导, 生成 code_map.md.

Archive root-level planning files (`task_plan.md`, `findings.md`, and `progress.md`) into a timestamped `.plannings/` directory for the current feature. Use when Codex should run `$archive-planning [feature-name]`, close out an active plan, or preserve completed planning context.

Write, debug, and validate ast-grep structural code search rules. Use for syntax-aware code search, AST pattern matching, structural refactor discovery, or searches plain text tools like rg can miss — functions with particular descendants, calls inside specific contexts, decorators, or other Tree-sitter-backed structures.

Audit and improve Claude Code CLAUDE.md guidance files, .claude/rules/ path-scoped rules, and companion code_map.md navigation maps. Use when the user asks to check, audit, optimize, or restructure CLAUDE.md, mentions nested CLAUDE.md or code_map.md, or says 优化 CLAUDE.md, 审计 CLAUDE.md, 生成 code_map (Claude). Not for trivial single-line edits the user has already fully specified.

Analyze a repository and current Codex environment, then recommend Codex CLI/App, AGENTS.md, skills, subagents, plugins, MCP servers, config/hooks, and OMX workflow improvements without modifying files. Use when the user asks to optimize Codex workflows or setup, recommend Codex automation, configure MCP/plugins/subagents, or says 优化 Codex 流程, Codex 工作流推荐, 给 Codex 配 MCP.

Turn vague or complex agent tasks into strong `/goal` commands for Claude Code and Codex — outcome, verification, constraints, boundaries, iteration policy, and stop conditions. Use for Codex/Claude Code goal instructions, Goal 指令, 目标指令, `/goal` prompts, 中文 Goal 模板, plan-to-goal interviews, or bounded agent work definitions.

Use when the user needs text or regex content search with ripgrep: composing rg commands, choosing flags, glob/type filtering, multiline or PCRE2 searches, pipeline output, grep-to-rg migration, or diagnosing why rg missed a file (gitignore, hidden, binary defaults). Not for syntax-aware structural queries (ast-grep) or semantic renames/references (language tooling).

Audit and safely clean Windows dev-process buildup and UWP background-task pileups — stale node/npm/cmd/pwsh trees (leaked Playwright MCP workers, dev servers, IDE services) and backgroundTaskHost.exe pileups. Use when Task Manager is full of node/npm noise, or for Windows 进程堆积, 清理残留开发进程, 查杀泄漏的 Playwright MCP, backgroundTaskHost 堆积.

Structured code review across correctness, security, performance, readability, testing, and architecture. Use whenever the user asks to review a PR, inspect git changes before merge, audit a directory or file set, prepare merge feedback, or do code review / PR review / CR / 代码审查. Output language follows the discussion language (Chinese or English).

Run a code quality review focused on maintainability, structure, abstraction quality, and refactoring opportunities. Use when the user asks for code quality review, maintainability review, PR code quality feedback, 代码质量审查, 可维护性审查, 架构质量审查, or review comments about code structure. Not for pure security review, formatting-only review, or performance profiling.

Implement safe, behavior-preserving code refactors after inspecting the existing project. Use when the user asks to refactor, split large files, extract methods, reduce duplication, rename, clean dead code, or says 重构代码, 拆分模块, 提取方法, 优化命名, 优化注释, 删除未调用代码. For broad requests, plan safe slices and wait for approval; for narrow ones, implement the smallest verifiable slice.

Use only when the user explicitly asks for swarm, subagents, parallel agents, dynamic workflow, multi-agent orchestration, 多智能体编排, or when the task truly needs coordinated research plus implementation plus review plus verification packets. Not for ordinary code review, planning-only work, single-line bugfixes, or routine audits unless orchestration is requested.

Challenge ideas, plans, products, pricing, pitch narratives, and major decisions with a no-flattery adversarial review. Use for 泼冷水, 挑刺, 给我泼盆冷水, 魔鬼代言人, devil's advocate, challenge assumptions, 哪里会崩, hidden assumptions, failure modes, kill criteria. Not for ordinary code review unless the user explicitly asks to challenge assumptions.

Use when the user explicitly asks to think bigger, open up the design space, challenge conservative design, escape local-detail fixation, or make a bold high-level product or architecture direction call. For strategic reframing — not ordinary code review, PRD writing, or adversarial risk review.

Ground an ambitious proposal into the smallest verifiable first move with stop rules — pressure-test feasibility, avoid over-grand designs, make a bold direction executable. Trigger on 落地, 先落地, 别太飘, 收一收, 可执行, 可验证, 止损, or follow-ups after geju-style big-picture thinking. Not for ordinary code review or implementation unless asked to ground or shrink the plan.

Write a handoff.md that lets a fresh session resume unfinished work cold. Use when compacting before auto-compaction, handing off an unfinished task, switching topics, or proactively when context usage is high. Triggers include 上下文压缩, 交接文档, 接力, 续接会话, 明早接着干, handoff, fresh session. Do NOT use for work that fits in one commit body, pure Q&A, or human-facing retrospectives.

Create self-contained HTML artifacts (single-file by default, split bundles when oversized) for complex, reviewable, or shareable work outputs. Use for an HTML artifact, browser-viewable report, implementation plan, PR/code-review report, dashboard, or mini deck, or for long Markdown-like outputs needing navigation or diagrams. Not for short answers, commit messages, or production UI.

Maintain a live implementation-notes.md while implementing a multi-step spec, PRD, design doc, or approved plan — capturing design decisions, intentional deviations, rejected alternatives, and open questions. Start before the first edit. Triggers include 实施记录, 决策日志, 边写边记, 按这个 spec 实现, decision log. Do NOT use for single-file fixes, pure investigation, or work without a written spec.

Turn an idea into an approved implementation plan before coding — plan-first brainstorming. Use when the user wants to brainstorm, design, scope, stress-test, or plan a feature/spec. Codex native Plan mode stays read-only with a chat-only final plan; Claude Code Plan mode writes only after approval exits Plan mode; writable/default mode saves Markdown plans under .plannings/YYYY-MM-DD-feature-slug.md.

Diagnose a task before execution when the user may not yet know how to define success. Use to clarify an ambiguous or unfamiliar task, define what good means, turn vague intent into an actionable brief, or when the user says 澄清任务, 先诊断再做, 理清需求, 帮我定义成功标准. Do NOT use for fully-specified small tasks (execute directly), feature-to-plan brainstorming (spark), or challenging an existing plan (cold-shower).

Create technical diagrams — architecture, workflow, sequence, data-flow, lifecycle/state — as standalone HTML files with SVG graphics, dark/light theme toggle, and built-in image/SVG export. Accepts plain language or pasted Mermaid code, laid out from scratch in archify style. Also use to convert or beautify a Mermaid diagram.

Modify the Beautiful Mermaid live editor itself rather than writing ordinary Mermaid diagrams. Use when the task mentions the Beautiful Mermaid repo, editor.ts, generated editor.html, config panel, themes or dark mode, zoom, PNG/SVG export, sample presets, or renderer wiring. Not for generic Mermaid syntax help.

智能招投标文件编写专家，覆盖工程咨询、建筑设计、市政、IT、软件开发、货物与服务采购等各类招投标，工程建设类为深度强项。能解析招标文件、提取评分标准与废标条款、制定投标策略、撰写技术标与商务标、做逐条响应与合规性及废标风险审核。当用户提到标书、投标、招标、技术标、商务标、评分标准提取、废标风险、逐条响应、偏离表、政府采购、tender、RFP response 时使用。不适用于通用商务写作、营销文案、学术论文或与招投标无关的文档。

Write or update technical documentation grounded in the real codebase. Use when the user asks for README, API docs, architecture guides, user guides, CONTRIBUTING docs, migration notes, or JSDoc/code comments, or wants technical docs rewritten into natural Chinese with correct terminology. Not for general prose, marketing or social copy, or non-technical localization.

Chinese public-writing editor for AI/tech posts, X/Twitter threads, product notes, and public technical essays. Use for 去AI味, 改得像本人, 写推特post, 精修中文AI技术文章 — removes AI-flavored shells while preserving facts, judgment, technical terms, and author voice. Not for academic papers, codebase docs, bid documents, or AI-detector evasion.

Author Typst slide decks with the Touying package. Use when creating or editing .typ presentations: applying or building Touying themes (e.g. metropolis), turning headings or #slide into slides, incremental reveals (#pause, dynamic content), speaker notes via pdfpc, and page layout config. Not for non-Typst decks (HTML, PPT, reveal.js, Marp) or ordinary non-presentation Typst documents.

Safely orchestrate Conventional Commits for staged Git changes, or all working-tree changes when the user explicitly asks to include everything. Use when asked to write a commit message, split staged or working-tree changes, organize a messy index, or draft commit text without pushing. Auto-detects commit language; explicit phrases like 用中文提交 or 请使用中文拆分提交所有的改动 override.

Multi-source deep research for current-topic investigation, comparison, and cited report writing. Use when the user asks to research, compare, or deep-dive a topic with current web sources and citations — research, deep dive, latest, 调研, 做个深度研究, 带来源总结. Not for offline-only codebase questions or when web access is forbidden.

Register-aware academic language polisher for English journal articles and Chinese doctoral dissertations: removes AI-writing tells while keeping academic norms. Use whenever the user wants to polish, 降AI味, 润色, or norm-check an academic draft, paper section, abstract, or 学位论文 paragraph. Polishes the author's own drafts only — not for laundering generated text past Turnitin or 知网 AIGC detection.

文献深度解读助手，像研究生导师一样交互式解读 Zotero 库中的学术论文（计算机/深度学习/自动化方向，个人向）。当用户提供文献题目、DOI、PDF 或要求解读某篇论文时触发；通过 Zotero MCP 优先获取全文，按用户意图自动选择快速筛选、导师深读或研究复盘模式。多篇论文综合、对比、找研究空白或批量规范化时改用 paper-workbench。

Generate publication-quality matplotlib figures in real academic paper styles. Two modes: from-data (pick a pre-built catalog style, fill in your numbers) and from-image (reproduce an uploaded paper figure as a matplotlib script). Use for "把数据画成论文图", "用某风格画我的数据", "复现这张图", or a named catalog style. Drawing only, no paper reading; route journal-compliance figures (IEEE/Elsevier/Nature specs) to academic-figure.

Researcher-profile-driven paper intake and literature workbench for academic workflows. Use whenever the user wants to skim, deep-read, card, compare, synthesize, map research gaps, or build a literature review from papers, arXiv/AlphaXiv links, DOIs, PDFs, or existing workbench artifacts. Trigger even when the user only says “精读这篇”, “整合这几篇”, “找研究空白”, or “搭综述框架”.

Multi-perspective roundtable discussion — a truth-seeking moderator plus a small set of representative real figures explore one topic through disciplined debate. Use for 圆桌讨论, roundtable, 多人物观点碰撞, 让几位思想家讨论, 结构化辩论探索. Not for ordinary argumentative writing, pros/cons lists, or single-speaker explanation.

Use when a coding agent needs to run Python code, modules, one-liners, tools, tests, or standalone scripts through uv, or create and maintain PEP 723 scripts with uv init/add/remove --script. Avoid direct python/python3 shell entrypoints.

将 UI 截图、设计稿、参考图复刻为可点击的前端/App demo:拆分代码渲染 UI 与必须生成的位图资产，生成提示词并把生成图接回页面。also use for image to UI, UI screenshot to code, clickable app demo, mobile prototype, iOS preview, high-fidelity UI recreation。涉及生图时优先项目指定 image2 入口，失败再走已登记的 OpenRouter ICU gpt-image-2 备案通道并标明实际通道；不要用 imagegen 或其他未指定工具替代。要求做成 App/手机/iOS 预览时，交付带 iOS 外边框的可点击预览与截图验真。

Address GitHub PR review comments and actionable review threads with GitHub CLI. Use when triaging reviewer feedback, summarizing unresolved PR comments, or applying selected fixes on a pull request.

Initialize GitHub repository configuration from vetted upstream templates. Use when setting up repository automation, issue and PR templates, CI workflows, or baseline GitHub project files for a new or existing repo.

Debug and fix failing GitHub PR checks with GitHub CLI. Use when GitHub Actions checks fail, when a PR has broken status checks, or when you need local reproduction commands for CI failures.