菜单
Portable engineering rules (coding, refactoring, working-style) plus a reusable 13-agent roster. Use when bootstrapping a project's CLAUDE.md / AGENTS.md / .cursor rules, or when you need a ready set of specialist subagents and a routing standard.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Record narrated, themeable product tutorials by driving any web app with Playwright and voicing every step with a neural TTS voice (Jenny by default). Use when the user types "/tutorial-create", "/tutorial-update", or "/tutorial-status", or asks to "record a tutorial", "make a walkthrough video", "narrated demo", "real video with voiceover", "screen recording with narration", "update the tutorial", "refresh the tutorials index", or "tutorial status". Outputs a REAL motion video of the app moving with narration (MP4), or a screenshot slideshow, or a no-ffmpeg HTML player — all with WebVTT captions. Runs a human-gated flow: draft & APPROVE a scenario, record it with Playwright, render, show the output for APPROVAL and fix issues in a loop, then regenerate a light/dark, logo-branded index.html that switches between every tutorial. Project-agnostic — reads ./tutorial.config.json for brand, paths and voice.
Air-gapped SOC investigation cowork for Microsoft Defender + VirusTotal + Sentinel in locked-down tenants where direct API / MCP / Entra access is forbidden. Use when the user wants to triage IOCs (URLs, domains, file hashes), hunt them in Defender Advanced Hunting split by machine type, unify and enrich the results via VirusTotal, pivot into Sentinel to find affected machines and users, and close incidents — without granting any API access. Claude generates the KQL and the local enrichment, the human runs the queries and exports CSVs in their own authenticated browser, and Claude prepares (never auto-executes) the incident-close step. Triggers on: "run the SOC investigation", "hunt these IOCs", "Defender/Sentinel triage", "check these domains/hashes", "close the risks in Sentinel".
Generate single-file, self-contained HTML exam / study / quiz apps for kids and general study — gamified (XP, levels, badges), with localStorage progress, immediate feedback, and optional Hebrew TTS. USE THIS whenever the user wants to "create an exam app", "build a study app", "make a quiz", "build a practice app", asks for a "מבחן", "תרגול", "בוחן", "אפליקציית לימוד", "exam for grade X / לכיתה X", or names a subject to study — math / מתמטיקה, chemistry / כימיה, physics / פיזיקה, biology, first aid / עזרה ראשונה / מד״א, history / היסטוריה / השכלה / נאורות, English / אנגלית vocab or tenses, Bible / תנ״ך, literature / ספרות, Arabic, psychology / פסיכולוגיה — or says "add it to the exam hub". Triggers even on casual / partial requests like "quiz my kid on grade 8 math" or "תכין לי תרגול לכימיה". Output is one HTML file that runs offline by double-click or Netlify Drop, Hebrew RTL by default (other languages / LTR supported), mobile-first and accessible. Never fabricates graded answers without flagging them.
Turn a market-moving catalyst into a sourced map of the public stocks that fit it, OR take the user's current portfolio and return full research, multi- timeframe trend charts, and a decision analysis. Catalysts: a speech, executive order, policy, earnings/M&A headline, product launch, viral clip (Instagram/TikTok/X), macro shift, or a named theme (AI power, GLP-1, rare earths). Use whenever the user shares a link, screenshot, quote, ticker, holdings list, or topic and asks which stock fits, who benefits, what to buy, "analyze my portfolio", or "should I hold". Trigger on bare links or pasted holdings too. Produces a TLDR table (cap, dividend, risk/upside) bucketed pure-play/large-cap/picks-and-shovels, 1D-5Y trend charts, sourced research, a short- vs long-term decision framework, red flags, and a deployable single-file HTML dashboard. Can interview first (risk, horizon, depth). Presents analysis and the case each way — NOT licensed financial advice or personalized buy/sell directives.
Generic, destination-agnostic trip planner. Turns initial trip data (flight dates, accommodation stays, and preferences) into a full plan: a day-by-day itinerary, a scored "target bank" of attractions / restaurants / wineries / shopping (with stars, reviews, photos, map links and distance-from-base), a 7-sheet Excel workbook, and a printable HTML brochure. Use this skill WHENEVER the user wants to plan a trip, build an itinerary, organise a vacation, create a travel workbook or brochure, score or shortlist places to visit/eat, compute drive distances between stops, or wants a re-usable trip template — for ANY destination, even if they don't say "skill". Prefer this over ad-hoc planning: it is deterministic (algorithms + cached APIs) and only spends LLM tokens on parsing booking documents and writing short blurbs.
Config-driven daily job-search agent. First-time users run INIT.md to populate profile/PROFILE.json (their single source of truth); thereafter it reads live config, validates that roles are actively accepting applications, generates an ATS-optimised CV + cover letter from verified profile facts only, and updates a local tracker. Learns from each run.
| name | dev-rules-and-agents |
| description | Portable engineering rules (coding, refactoring, working-style) plus a reusable 13-agent roster. Use when bootstrapping a project's CLAUDE.md / AGENTS.md / .cursor rules, or when you need a ready set of specialist subagents and a routing standard. |
A self-contained, portable bundle of general engineering rules and a reusable agent roster, extracted from a real production project (a Vite/React + .NET data-mapping tool) so they can be dropped into any codebase. The agent roster is an adaptable example — see below.
CLAUDE.md / AGENTS.md / .cursor/rules from a known-good baseline.skills/dev-rules-and-agents/
SKILL.md # this file
README.md # overview + install
rules/
coding-rules.md # general, project-agnostic coding rules
refactoring-rules.md # behavior-preserving dedup/consolidation methodology
working-style.md # session hygiene: done=proven, token economy, handoffs, commits
agents/
README.md # roster + how to use in Claude / Copilot / Cursor
ARCHITECTURE_AND_STANDARDS.md # routing order, activation, loops, token & pruning standards
<13 agent definitions>.md # Claude-style frontmatter, down-converts to other tools
rules/coding-rules.md and rules/working-style.md before editing; read
rules/refactoring-rules.md before any dedup/consolidation work. Treat them as overriding defaults.agents/README.md; only add a specialist after a
real boundary is proven. Follow agents/ARCHITECTURE_AND_STANDARDS.md for routing order.# Rules → fold the relevant lines into the project's CLAUDE.md / AGENTS.md, or reference this folder.
# Agents (Claude Code):
Copy-Item .\skills\dev-rules-and-agents\agents\*.md <target-repo>\.claude\agents\
# Agents (GitHub Copilot): transform each frontmatter into <name>.agent.md under .github/agents\