magiccube/agentara/daily-hunt

Daily Hunt 🏹

A personalized daily tech discovery digest combining Product Hunt launches and GitHub Trending repositories, tailored to the user's interests and expertise.

Workflow

Step 1: Determine Date & Data Sources

1. Note today's date.
2. Product Hunt launches happen Mon–Fri (Pacific Time). If today is Saturday/Sunday, use the most recent Friday's leaderboard instead and note this in the output.
3. Construct the PH leaderboard URL: https://www.producthunt.com/leaderboard/daily/{YYYY}/{M}/{D}

Step 2: Fetch Product Hunt Data

Use web_search and web_fetch to gather PH daily leaderboard data:

1. Primary: Fetch https://www.producthunt.com/leaderboard/daily/{YYYY}/{M}/{D}
2. Backup: Search hunted.space history {month} {year} for structured launch data with upvotes
3. Backup 2: Fetch https://hunted.space/stats for recent daily stats
4. Try to get at least the top 5-10 launches with:

• Product name
• Tagline / one-line description
• Category tags
• Upvote count (if available)
• Comment count (if available)

1. For each product, construct its PH link: https://www.producthunt.com/posts/{product-slug}

• If the exact slug is unknown, link to the daily leaderboard page instead

Step 3: Fetch GitHub Trending Data

1. Primary: Use web_fetch on https://github.com/trending to get today's trending repos
2. Extract top 10 repositories with:

• Repository full name (owner/repo)
• Description
• Primary language
• Star count (total)
• Fork count
• Stars gained today

1. Construct links: https://github.com/{owner}/{repo}

Step 4: Personalize & Curate

Cross-reference the fetched data with the user's profile from memory (userMemories):

1. Identify relevance signals — Check each item against the user's:

• Tech stack (languages, frameworks, tools)
• Work domain (e.g., medical tech, AI agents, developer tools)
• Side projects and interests (e.g., agent architecture, AIGC, photography)
• Favorite topics (e.g., open source, LangChain, MCP, TypeScript)

1. Mark noteworthy items with emoji indicators:

• ⭐ — Directly relevant to the user's tech stack or active projects
• 🔥 — Exceptionally hot / high engagement / viral
• 👀 — Interesting for the user's broader interests or domain
• No emoji — Standard inclusion, still worth knowing about

1. Preserve original ranking — Keep the exact order from the source (PH leaderboard rank, GitHub Trending page order). Do NOT re-sort by engagement or relevance. Only add emoji indicators (⭐🔥👀) to mark noteworthy items in-place.

Step 5: Format Output

Use the user's preferred language (check memory — if Chinese-speaking, use Chinese with full-width punctuation).

Output Structure

Use list format (not tables) for all items — this renders correctly on mobile and in IM apps.

# Daily Hunt 🏹 — {date}
 
> {One-sentence personalized summary of today's highlights}
 
## Product Hunt 热门发布
 
1. ⭐ **[Name](link)**
   tagline
   → personalized reason why it matters
 
2. **[Name](link)**
   tagline
   → reason
 
...
 
---
 
## GitHub Trending Top 10
 
1. 🔥 **[owner/repo](link)** · Python · ⭐31.4k · 🍴2.1k · +1,122 今日
   description
 
2. ⭐ **[owner/repo](link)** · TypeScript · ⭐8.9k · 🍴430 · +300 今日
   description
 
...
 
## 📌 昕哥精选
 
{2-3 sentences highlighting the most relevant items for this specific user,
explaining WHY these matter for their work/projects. Reference specific
projects or interests from memory.}

Key Rules

• Use lists, not tables — tables break on mobile and in most IM apps
• Product Hunt items MUST include: name, link, tagline, personalized "why care" note (prefixed with →)
• GitHub items MUST include: full repo name, link, language, star count, fork count, today's stars, description
• "Today's stars" (stars gained today) should be shown when available from the trending page
• If data for a source is partially unavailable, still show what you can and note the gap
• The "Personal Picks" section at the end is the differentiator — make it genuinely insightful
• Do NOT pad with generic descriptions — be specific about why each item matters to THIS user
• Use Chinese full-width punctuation (，、：！) for Chinese content
• First-level headings use # not ##
• Keep the "why care" note concise (one sentence max)