菜单
Get your personalized trend briefing based on profile domains.
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
Install hookify rules for automatic chess timer session management
Create a custom source adapter for pattern-radar
Capture thinking visually during brainstorming, architecture discussions, or explaining flows. Creates persistent Mermaid diagrams.
Automatically use Perplexity search for personalized context enrichment. Use when facing unfamiliar technologies, decision points, learning questions, or preference-dependent choices.
Guide for storing enriched memories that capture decisions, preferences, and context. Use when making significant decisions or learning user preferences.
Guided workflow for reverse engineering black-box systems. Use when a user wants to decode a defunct API, replicate a competitor's feature, understand unfamiliar code, or crack a data format.
| name | pattern-radar:digest |
| description | Get your personalized trend briefing based on profile domains. |
This skill dispatches a Haiku subagent to generate your personalized trend briefing. Main context receives a curated narrative; subagent does the analysis.
You (Opus) → Dispatch Haiku subagent → Subagent calls radar tools → Returns narrative
← Personalized briefing ← Raw signals processed
When user asks for trends, digest, or wants to know what's happening:
Step 1: Dispatch Digest Subagent
Use Task tool:
- subagent_type: "general-purpose"
- model: "haiku"
- prompt: "You are a personalized trend analyst. Generate a trend briefing for the user.
STEPS:
1. Call get_radar_digest() to fetch signals from HN and GitHub
2. Call get_intersections() to find domain overlaps
3. Call suggest_actions(focus: 'all') for actionable recommendations
OUTPUT FORMAT:
Return a 300-400 word personalized briefing with these sections:
## 🎯 Relevant to You
[2-3 patterns directly matching user's domains, with why they matter]
## 🌟 Wildcards (Serendipity)
[1-2 interesting signals OUTSIDE user's usual domains - unexpected discoveries]
## 💡 Suggested Actions
[2-3 specific actions: something to learn, something to build, something to explore]
Keep it conversational and actionable. Include specific links where helpful."
Step 2: Present the Briefing
Tell the user the subagent's narrative. Don't dump raw signal lists.
IMPORTANT: Always include the Wildcards section. The whole point of pattern-radar is surfacing unexpected connections.
If user explicitly asks for "focused mode" or "only my domains":
For quick topic scans, use tools directly (no subagent needed):
scan_trends(topic: "WebAssembly")
This returns raw signals - useful when user wants to explore a specific topic themselves.
Help users configure their radar:
configure_sources(domains: ["TypeScript", "AI", "distributed-systems"])
Or for source weights:
configure_sources(hackernews_weight: 1.5, github_weight: 1.0)
User: What's trending that I should know about?
You: Let me generate your personalized trend digest...
[Dispatch Haiku subagent]
Subagent returns:
## 🎯 Relevant to You
**Edge Runtime Performance** is heating up - 3 HN discussions and 2 trending repos
about V8 isolates and cold start optimization. Given your TypeScript + serverless
background, this could affect your Lambda work.
**AI Code Review Tools** are seeing a surge. GitHub Copilot competitors like
Cursor and Continue are trending. Worth watching for your code-review plugin work.
## 🌟 Wildcards
**Game Engine Renaissance** - Godot 4.3 just dropped and HN is buzzing. Not your
usual domain, but the architectural patterns (ECS, hot-reload) might inspire
plugin design.
## 💡 Suggested Actions
- **Learn:** Read the Cloudflare Workers V8 isolate blog post (linked in signals)
- **Build:** Try adding AI-assisted review to your code-review plugin
- **Explore:** Check out Godot's plugin architecture for inspiration
User: Just show me AI stuff, skip the wildcards
You: [Call scan_trends directly]
Here are the AI-specific signals...