About this project

This is a skills collection repository for AI Agents, authored by "Digital Life Kazik". Each Skill is a structured instruction set that Agents can load directly, following the open Agent Skills standard. Officially, 40+ agents supporting this standard can install them, including Claude Code, Codex, Qoder, Kimi Code, iFlow, CodeBuddy, and Cursor. Installation involves providing the link to the corresponding skill directory in the repository within a Skill-supporting tool, letting the Agent clone it itself; if the Agent doesn't support Skills, the SKILL.md can be downloaded as a project rules file or pasted directly into the conversation for execution. The repository currently contains six skills: 1. leader (Leadership): Does one thing only—turns vague requirements into a "Mission Task Document" for goal mode (e.g., Claude Code's /goal, Codex's goal mode), enabling the AI to execute long-term tasks independently. The documentation emphasizes that the constraints on "what not to do" (Harness) in the goal are more critical than the goal itself, and provides a "Seven Goal Questions" framework (purpose, completion state, evidence, anti-cheating, boundaries, trade-offs, unknowns). Output is pure Markdown. 2. storage-analyzer (Junk Cleanup): Triggers a full-disk scan with a single sentence, generating an interactive HTML report in the browser, including disk overview, Top 5 space consumers, cleanup priorities, and a green/yellow/red three-tier classification list. Green items are pure cache that can be cleaned with one click; yellow items contain user data and only offer opening or moving to Trash; red items are system or running-app data, explained but not deletable. Deletion requires clicking in the browser with secondary confirmation; the local service runs on 127.0.0.1 with a random port and token. The documentation states macOS has been fully tested, Windows code is ready but first-time use should be cautious. 3. aihot (AI HOT News Query): Lets Agents query daily AI reports and updates from aihot.virxact.com using natural language, without API keys or MCP configuration. Supports searching by date, category, time window, and keywords, viewing current hottest events, and syncing selected content locally. 4. neat-freak (Neat Freak): Runs /neat at session end to align this session's changes with project documentation, CLAUDE.md / AGENTS.md, and Agent memory across three layers, auditing whether rules were followed and outputting a change summary. The documentation notes a lightweight path for small projects, and all deletions only generate candidate lists requiring user confirmation. 5. hv-analysis (Horizontal-Vertical Analysis): Conducts dual-line research on products, companies, concepts, or people—vertical timelines and horizontal competitor comparisons—finally outputting a 10,000–30,000 character PDF research report. The documentation states it suits competitor research and systematic material preparation, not simple term explanations. 6. khazix-writer (Kazik Writing): A writing skill based on the author's personal WeChat official account style, including style rules, a four-layer self-check system, and an example library, explicitly listing cliché expressions it will reject. It's positioned as opinionated stylistic writing rather than general prose. The repository is open-sourced under the MIT license, allowing free use, modification, and redistribution. The README also includes links to several explanatory WeChat official account articles.