RonanCodes/skills/producthunt-scan

Product Hunt Scan

Three modes:

• Today feed — default, no topic — 50 freshest launches
• Topic filter — filter today/week feed by keyword match on title or tagline
• Category feed — scope to a PH category (e.g. tech, productivity, ai)

Product Hunt's public Atom feed is https://www.producthunt.com/feed — returns the current top products, no auth, atom-XML format.

Usage

/ro:producthunt-scan                                 # top 30 launches today
/ro:producthunt-scan "AI agent"                      # filter today's feed by keyword
/ro:producthunt-scan "markdown" --since week         # past week filtered
/ro:producthunt-scan --category ai --limit 50        # AI category, more items

Defaults

• Since: today
• Limit: 30
• Category: none (full feed)

Endpoints

# Today / main feed
curl -s "https://www.producthunt.com/feed"

# Category-scoped (replace <slug>)
curl -s "https://www.producthunt.com/feed?category=<slug>"

Common category slugs: tech, ai, productivity, design-tools, developer-tools, marketing, web3, e-commerce.

Parsing

The feed is Atom not RSS. Use Python's xml.etree.ElementTree with namespace http://www.w3.org/2005/Atom:

import xml.etree.ElementTree as ET
ns = {'a': 'http://www.w3.org/2005/Atom'}
tree = ET.parse('/tmp/ph.xml')
for e in tree.getroot().findall('a:entry', ns):
    title = e.findtext('a:title', default='', namespaces=ns)
    link = e.find('a:link', ns).get('href')
    updated = e.findtext('a:updated', default='', namespaces=ns)
    summary = e.findtext('a:summary', default='', namespaces=ns)
    # id → product slug
    # author → posted by

Note: the public feed does NOT expose upvote counts. If you need upvotes, the product page itself has them in JSON-LD (scrape the product URL), or use the GraphQL API (requires auth token — out of scope).

Topic filter

When a topic is given, fetch the feed and filter entries where title or summary matches the topic (case-insensitive substring). PH titles are often 1-2 words so be lenient — also match the product page content if the topic is broad.

For better matching, fetch each candidate product page and check the tagline + description:

curl -s "https://www.producthunt.com/products/<slug>" | \
  python3 -c "
import sys, re
html = sys.stdin.read()
# tagline is in <meta property='og:description'>
m = re.search(r'<meta property=\"og:description\" content=\"([^\"]+)\"', html)
print(m.group(1) if m else '')
"

Rate-limit yourself — don't hammer product pages. Cap at ~20 per scan.

Output

# Product Hunt  (since: <today|week>, category: <all|slug>, filter: "<topic or none>")

1. **<Product name>** — <tagline / og:description>
   <https://www.producthunt.com/products/<slug>>
   Launched: <date>  |  Category: <inferred from URL path or description>

2. ...

## Themes
- <1-line pattern spotted across today's launches>
- <another>

Discover mode (for trend-scan)

Called by /ro:trend-scan --discover. Pull the main feed for today, cluster by theme (keyword extraction on titles + summaries), return the top themes with 2+ product mentions.

Error handling

• Empty feed / parse error — PH sometimes returns an HTML error page instead of XML. Check response is application/atom+xml before parsing.
• Topic yields zero results — widen to week window (--since week) automatically and note the fallback.
• Product page scrape blocked — PH throttles aggressive scraping. Back off to 1 req/2s; if still blocked, skip the per-product enrichment.

See also

• hn-scan, reddit-scan, x-scan, youtube-scan — sibling source scanners.
• trend-scan — upstream orchestrator.
• Product Hunt's GraphQL API (https://api.producthunt.com/v2/docs) — if upvote counts matter, upgrade to authenticated calls.