Cto Intelligence

CTO Intelligence Briefing

Set up recurring monitoring for a technology domain and deliver formatted briefings. Use when asked to set up daily/weekly monitoring, a briefing, a cron job for news, or "keep me updated" about a tech domain.

Structure

1. Identify the Domain

Clarify scope:

• Core domain (e.g., agentic AI, cybersecurity, LLM ops)
• Sub-categories to monitor
• Delivery preference (Telegram DM, email, daily/weekly)
• Time preference (morning, evening, specific hour)

2. Search Strategy

For each sub-category, define 2-3 search queries:

Category: [name]
Queries:
  - "latest [category] news 2026"
  - "[category] update release 2026"
  - "new [category] tool framework 2026"

Available Search Methods

Two-Pass Search Workflow

Pass 1 — HN Algolia (parallel broad queries): Use curl to search_by_date endpoint. Filter by created_at_i timestamp in Python. No auth required.

Pass 2 — DDG lite for gaps: Run DDG lite batch for categories with thin coverage. Rate limit: 2 seconds between queries.

If both passes return empty, the topic had no news that day — record as quiet and move on.

Cron-Compatible Search Methods

In cron context, execute_code and browser_navigate are blocked. Use terminal-only:

# HN Algolia API via curl
curl -s 'https://hn.algolia.com/api/v1/search_by_date?query=KEYWORD&tags=story&hitsPerPage=30' --max-time 15 | python3 -c "
import json, sys, time
data = json.load(sys.stdin)
now = time.time()
for h in data.get('hits', []):
    ts = h.get('created_at_i', 0)
    age_hours = (now - ts) / 3600
    if age_hours > 48: continue
    title = h.get('title', '')[:120]
    pts = h.get('points', 0)
    url = h.get('url', '') or 'https://news.ycombinator.com/item?id=' + str(h.get('objectID', ''))
    print(f'{age_hours:5.1f}h | {pts:>4}pts | {title}')
    print(f'       {url}')
"

# DDG Lite via Python in terminal
python3 -c "
import urllib.request, urllib.parse, re, time

def ddg_search(query):
    url = 'https://lite.duckduckgo.com/lite/'
    data = urllib.parse.urlencode({'q': query}).encode()
    req = urllib.request.Request(url, data=data, headers={'User-Agent': 'Mozilla/5.0'})
    resp = urllib.request.urlopen(req, timeout=10)
    html = resp.read().decode('utf-8', errors='replace')
    results = re.findall(r'<a[^>]*href=\"(https?://[^\"]+)\"[^>]*>(.*?)</a>', html)
    return [(t.strip(), u) for t, u in results if t.strip() and 'duckduckgo' not in u.lower()][:5]

for q in ['query 1', 'query 2', 'query 3']:
    print(f'=== {q} ===')
    for title, url in ddg_search(q):
        print(f'  {title[:100]} -> {url}')
    time.sleep(2)
"

3. Build the Cron Job

Use cronjob tool with:

The prompt must be self-contained. Include actual search methodology, not just categories.

4. Post-Delivery Workflow

When user asks to act on a briefing:

1. Categorize findings: 🔴 Check (CVEs), 🟢 Install (tools), 📡 Monitor (trends), ℹ️ Note (no action)
2. Evaluate each tool/project: Visit GitHub, check stars, commits, releases, dependencies
3. Verify security claims across fleet: Check if affected packages are installed on all nodes
4. Install selected tools: Standard workflow (clone/build, smoke test, save knowledge doc)
5. Save to knowledge store: Write findings to a dated knowledge file

5. Pitfalls

• Subagents hallucinate release versions, CVEs, and dates — always verify.
• HN Algolia uses search_by_date endpoint, NOT search?sort=byDate (returns 400).
• execute_code and patch are blocked in cron context — use terminal with Python or curl.
• DDG lite can return 0 results — pivot to HN Algolia when that happens.
• printf breaks with emoji and % characters — use Python with open for file appends.
• Always check publication dates — stale articles can appear as top results.
• Use [SILENT] (exact, no other text) to suppress delivery when nothing new to report.