Hackernews Intel Skill for Claude

Hacker News Intel

Monitor Hacker News for keywords. On each run, fetch new posts via the HN Algolia API, deduplicate against a local SQLite cache, and send Slack alerts for unseen matches. Run manually or schedule via cron or GitHub Actions.

Step 1: Check Setup

1a. Verify required environment variables

Check that these are set:

• HN_KEYWORDS (required): comma-separated list of keywords to monitor
• SLACK_WEBHOOK (required): Slack Incoming Webhook URL for alerts

If either is missing, stop and tell the user:

• HN_KEYWORDS: list the topics you want to monitor, comma-separated. Example: claude code,LLM agents,deno runtime
• SLACK_WEBHOOK: create an Incoming Webhook at https://api.slack.com/apps. Select your workspace, enable Incoming Webhooks, and copy the URL.

1b. Verify script dependencies

ls /Users/ksd/Desktop/Varnan_skills/hackernews-intel/node_modules/better-sqlite3 2>/dev/null

If missing:

cd /Users/ksd/Desktop/Varnan_skills/hackernews-intel && npm install

Step 2: Run the Monitor

cd /Users/ksd/Desktop/Varnan_skills/hackernews-intel && node scripts/monitor-hn.js

Flags:

• --dry-run: print matches to stdout without sending any Slack alerts. Use this to preview what would be alerted before committing.
• --days=N: on the first run, look back N days (default: 1). Only applies when the SQLite cache is empty.
• --reset: clear the cache and start fresh. Use when you change your keyword list significantly.

What the script does:

• Reads HN_KEYWORDS from environment, splits by comma, trims whitespace
• Opens (or creates) a SQLite database at the path in HN_DB_PATH (default: ./hn-intel.db)
• For each keyword, calls https://hn.algolia.com/api/v1/search_by_date with numericFilters=created_at_i>lastSeen
• For each result, checks if the objectID is already in seen_posts — skips if yes
• If HN_MIN_POINTS is set, skips posts below that threshold
• For new matching posts, sends a Slack alert and inserts the objectID into the cache
• Updates poll_log with the run summary
• Prints a summary to stdout

Step 3: Review Results

After the script runs, read its stdout output. It will report:

Run complete.
  Keywords checked: N
  Posts found: N
  Posts alerted: N
  Posts skipped (already seen): N
  Errors: N

If Posts found: 0 and you expect results, check:

• Are the keywords specific enough? Very broad terms (e.g. "AI") may return 0 results if HN Algolia returns too many and the time window is narrow
• Is the --days window large enough for the first run? Try --days=7
• Is the DB cache too aggressive? Try --reset to clear it and rerun

If Posts alerted: 0 but Posts found: N, the deduplication is working — those posts were seen in a previous run.

Step 4: Schedule the Monitor

Option A: cron (macOS/Linux)

Add to crontab to run every 4 hours:

crontab -e

Add this line (adjust the path to match your install):

0 */4 * * * cd /Users/ksd/Desktop/Varnan_skills/hackernews-intel && node scripts/monitor-hn.js >> /tmp/hn-intel.log 2>&1

Option B: GitHub Actions (recommended for teams)

Create .github/workflows/hn-intel.yml in any repo:

name: HN Intel Monitor
on:
  schedule:
    - cron: '0 */4 * * *'
  workflow_dispatch:

jobs:
  monitor:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '20'
      - name: Install dependencies
        run: npm install
        working-directory: hackernews-intel
      - name: Run monitor
        run: node scripts/monitor-hn.js
        working-directory: hackernews-intel
        env:
          HN_KEYWORDS: ${{ secrets.HN_KEYWORDS }}
          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
          HN_MIN_POINTS: ${{ secrets.HN_MIN_POINTS }}

Add HN_KEYWORDS, SLACK_WEBHOOK, and optionally HN_MIN_POINTS as repository secrets.

Note: GitHub Actions does not persist files between runs, so the SQLite cache resets each run. This means every run will find all posts within the default 1-day lookback window again. For persistent dedup on GitHub Actions, store the DB in a persistent location (S3, artifact cache, or a dedicated branch).

Step 5: Tune Keywords and Thresholds

Adding or changing keywords:

Update HN_KEYWORDS in your .env or environment. No restart needed — the script reads it fresh each run.

Setting a minimum points threshold:

Set HN_MIN_POINTS=10 to only alert on posts with 10 or more points. This cuts noise for broad keywords.

Including comments (not just stories):

Set HN_INCLUDE_COMMENTS=true to also monitor comments mentioning the keyword. Off by default — comments generate significantly more volume.

Resetting the cache:

Run node scripts/monitor-hn.js --reset to clear seen_posts and poll_log. The next run fetches fresh from the lookback window.

Testing a new keyword:

Run node scripts/monitor-hn.js --dry-run --days=7 to see the last 7 days of matching posts without alerting anyone.