khazix-skills
Five practical skills for disk cleanup, AI news, document alignment, horizontal and vertical analysis, and writing.
Suggested prompts
About this skill
Khazix Skills
A set of AI Skills used daily by Khazix. Each Skill is a structured instruction set that can be directly loaded by agents following the Agent Skills open standard, supported by Claude Code, Codex, Qoder, Kimi Code, iFlow, CodeBuddy, Cursor, and 40+ other tools. Includes 6 practical Skills covering goal definition, disk cleanup, news query, project wrap-up, deep research, and writing style.
leader
Helps turn a vague idea into a clear goal task sheet that an AI can execute autonomously for hours until completion.
Use Cases
- When you need the AI to independently run a long task
- When you want to define boundaries and rules before starting to avoid drifting off
Core Capabilities
- Guides you through a series of questions to define the goal together
- Generates a task sheet with 7 key elements (purpose, completion criteria, evidence, anti-cheating, boundaries, trade-offs, handling unknowns)
- Emphasizes constraints (Harness), clearly stating what must not be done
How It Works
- You say a rough idea
- The AI first investigates the project code
- It asks you at most 5 key decision questions
- Generates a Markdown task sheet
What You'll Get
A plain Markdown task sheet that can be pasted into goal mode for the AI to execute.
Usage Examples
帮我给 agent 写个目标
帮我详细拆一下这个目标
写个 goal 提示词
让 agent 自己跑这个项目
storage-analyzer
Scans your Mac/Windows disk with a single command and generates an interactive HTML report with three-tier color-coded cleanup suggestions.
Use Cases
- Running low on disk space and wanting to know which files can be safely removed
- Wanting a clear overview of storage usage
Core Capabilities
- Full-disk scan to identify large files and caches
- Generates an HTML report with disk overview, top 5 space consumers, and 🟢🟡🔴 classified list
- Green items (pure cache/temp files) can be deleted in one click; yellow items (containing user data) can be moved to trash; red items (system/app core) are view-only, no delete button
- Runs locally; deletion requires clicking a button in the browser and a second confirmation dialog
How It Works
- You say "help me check storage", the AI scans the disk
- Analyzes results and generates an HTML report
- Starts a local web service (127.0.0.1 + random port + token)
- Opens the report in your browser for you to review and decide actions
What You'll Get
An interactive HTML cleanup report to safely guide disk cleanup.
Usage Examples
帮我看看存储
C 盘满了
清理一下磁盘
看下电脑空间
storage analysis
Notes
- The scan is read-only; no files are automatically deleted
- Deletion must be manually confirmed by you in the browser
- Fully tested on macOS; Windows code supports multiple drives, but first-time use is advised to be cautious
- The report shows specific path, type, impact of deletion, and recommended action for each item
aihot
Query aihot.virxact.com for AI daily news and trends using natural language, no API key required.
Use Cases
- Quickly catching up on the latest AI hot topics
- Searching AI news by category, time range, or keyword
Core Capabilities
- Fetch today's or a specific date's AI HOT daily digest
- View the current hottest events (sorted by heat)
- Browse entries by category (model/product/industry/paper/tips)
- Filter by time window (past 24 hours, past 7 days)
- Search by keyword/company/topic (e.g., "OpenAI", "Sora", "RAG paper")
- Sync all current highlights to local, then only receive updates
How It Works
- You speak your query in natural language
- The agent calls the backend service via the Skill
- Results are returned; you can ask follow-up questions
What You'll Get
AI news content matching your query.
Usage Examples
今天 AI 圈有什么新东西
现在 AI 圈最热的事件是什么
看一下 5 月 6 号的 AI 日报
最近一周的 AI 论文
最近 OpenAI 有什么发布
把 AI HOT 当前全部精选同步到本地
aihot.virxact.com · Integration guide
neat-freak
After each task in the agent, run /neat to align project docs, CLAUDE.md/AGENTS.md, and agent memory, audit rule execution, and produce a change summary.
Use Cases
- After finishing a development task in an agent, to keep docs and memory in sync with code
- Before handing off a project, to consolidate knowledge into docs and rules
Core Capabilities
- Checks files modified in the session against project documentation (README, docs/), AI rule files (CLAUDE.md/AGENTS.md), and agent memory
- Flags outdated content and suggests updates
- Audits whether project rules were correctly followed (e.g., referenced paths still exist)
- Lightweight path for small projects: auto-create a minimal AI rule file, list residual files for cleanup
- Never deletes autonomously: all deletions are presented as a checklist for your confirmation; machine-generated memory is read-only by default
How It Works
- You input
/neator a trigger phrase - The AI analyzes the diff between code changes and doc/memory state
- Generates a change summary and an update checklist
- You confirm, then the AI updates documents and memory
What You'll Get
A project knowledge synchronization report, ensuring the next AI session works with up-to-date information.
Usage Examples
/neat
跑一下洁癖
把文档和记忆整理一下
新人接手，帮我做个 clean handoff
Notes
- This Skill only handles knowledge alignment, not pure code tasks or data wrangling
- All deletion suggestions require your manual confirmation
hv-analysis
Conduct deep research on any product/company/concept/person and generate a PDF research report of 10,000–30,000 words.
Use Cases
- Competitive analysis, exploring a new domain
- Systematic material prep before writing
- Understanding a complex topic from scratch
Not Suitable For
- Looking up a term definition (use normal chat)
- Writing WeChat public account articles (use khazix-writer instead)
Core Capabilities
- Tracks the subject's historical development (vertical depth)
- Compares key competitors at each stage (horizontal breadth)
- Cross-analyzes to produce independent insights
How It Works
- Provide a subject name or description
- The AI simultaneously gathers data along vertical and horizontal axes
- Cross-analyzes and generates a PDF report
What You'll Get
A well-formatted PDF research report, 10,000–30,000 words.
Usage Examples
研究一下 Cursor 这家公司
帮我做个竞品分析
这个产品到底是怎么回事
帮我做个 deep research
khazix-writer
Write long-form articles in the style of Khazix (from the WeChat channel "数字生命卡兹克").
Use Cases
- You've read Khazix's articles and like the style; you want your AI to write in that tone
- Provide material (PDF, voice-to-text, a news link) and get a WeChat article written
Not Suitable For
- You want generic good writing, and dislike the specific style that rejects words like "赋能", "抓手", "闭环" etc.
- Your target audience prefers formulaic or template expressions
Core Capabilities
- Built-in complete writing style rules (rhythm, narration, judgement, rhetoric)
- Four-layer self-check (structure, rhythm, content, wording)
- A set of style examples for the AI to reference
- Will refuse to use certain words (e.g., "赋能", "首先...其次", "在当今 AI 快速发展的时代" etc.)
How It Works
- Provide material or a writing instruction
- The AI generates an article following the style rules
- Output after self-check
What You'll Get
A long-form article in Khazix's writing style.
Usage Examples
帮我写篇文章
把这个素材写成长文
按我的风格写一下
帮我续写