competitor-monitoring-system

name	competitor-monitoring-system
description	Stand up and run ongoing competitive-intelligence monitoring for a client — maintain a watchlist, establish a baseline, then track competitor content, ads, reviews, social, and product moves on a cadence (content/social/Reddit weekly, ads bi-weekly, reviews monthly, full re-baseline quarterly). Diffs each cycle vs stored snapshots and produces an intelligence report with recommended actions behind human checkpoints.
metadata	{"version":"1.0.1","category":"competitive-intel","type":"playbook"}

Competitor Monitoring System

Playbook: a stateful, scheduled orchestration over the competitor-intel and competitive-pricing-intel engines plus per-channel collectors. Durable snapshot history is what makes diffs meaningful. You, the agent, classify change significance, write the report, and gate actions behind human approval.

Sub-skills this chains

• competitor-intel — per-competitor baseline + quarterly refresh (run its scripts/flow).
• competitive-pricing-intel — pricing-change detection leg.
• Per-channel collectors bundled here: fetch_pages.py (content), fetch_feed.py (blogs), hn_fetch.py (Hacker News), render_page.mjs (ads/reviews/social), wayback_fetch.py (pricing history), snapshot_store.py (cross-cycle diffing).

When to use

• "Set up competitor monitoring for [client]." / "Track what [competitors] are doing."
• "Monitor [competitor] content and ads."

How to run

1. Define the watchlist

Per competitor capture: {name, url, founder_linkedin?, blog_url?, review_urls?, ad_library_urls?}. Store it as ${WORKSPACE}/supabase/watchlist.csv (or Airtable/Supabase when configured). This is editable by the human.

2. Initial baseline (per competitor)

Run the competitor-intel engine per competitor (its fetch_pages.py / fetch_feed.py / render_page.mjs), add a current-ad scrape and latest review pull, then persist each competitor's tracked fields as the baseline:

python3 ${SKILL_DIR}/scripts/snapshot_store.py save --entity acme \
  --input ${WORKSPACE}/acme_baseline.json --store ${WORKSPACE}/supabase/monitor_history.csv

3. Configure cadence

Schedule recurring runs via the host platform cron / monitoring routine, one per cadence: content weekly, social weekly, Reddit/HN weekly, ads bi-weekly, reviews monthly, full re-baseline quarterly.

4. Run a monitoring cycle (per channel)

# Content (weekly): refetch blogs + pages
python3 ${SKILL_DIR}/scripts/fetch_feed.py  --url https://acme.com/blog --output ${WORKSPACE}/acme_blog.json
python3 ${SKILL_DIR}/scripts/fetch_pages.py --url https://acme.com https://acme.com/product --output ${WORKSPACE}/acme_pages.json
# Hacker News (weekly)
python3 ${SKILL_DIR}/scripts/hn_fetch.py    --query "acme" --days 7 --output ${WORKSPACE}/acme_hn.json
# Ads (bi-weekly) + Reviews (monthly): render the libraries / review pages
node ${SKILL_DIR}/scripts/render_page.mjs --url "<meta-ad-library-or-g2-url>" --wait 6000 --output ${WORKSPACE}/acme_ads.json
# Pricing (monthly): Wayback history for the pricing leg
python3 ${SKILL_DIR}/scripts/wayback_fetch.py --url https://acme.com/pricing --output ${WORKSPACE}/acme_pricing_hist.json

Social (weekly): render_page.mjs for LinkedIn founder posts / X, or PHANTOMBUSTER_API_KEY (LinkedIn) / APIFY_API_TOKEN (X/Reddit at depth).

5. Diff vs previous + flag significant changes

python3 ${SKILL_DIR}/scripts/snapshot_store.py diff --entity acme \
  --input ${WORKSPACE}/acme_cycle.json --store ${WORKSPACE}/supabase/monitor_history.csv
python3 ${SKILL_DIR}/scripts/snapshot_store.py save --entity acme \
  --input ${WORKSPACE}/acme_cycle.json --store ${WORKSPACE}/supabase/monitor_history.csv

From the added/removed/changed output, flag: new features/pricing, content targeting your keywords, negative review trends (poaching opportunity), new ad campaigns, exec strategy statements. first_run: true = baseline established this cycle.

6. Produce the intelligence report (you, the agent)

Write "Competitor Intelligence — [Client] — Week of [Date]": key changes → recommended actions → per-competitor detailed findings. Deliver to the client's intelligence area (workspace file / channel attachment; SLACK_*/DISCORD_*/TELEGRAM_* if configured).

7. Human checkpoints

Post the watchlist/plan after setup and the recommended actions after each report and wait for approval before executing or routing to outreach skills.

Outputs

• ${WORKSPACE}/supabase/watchlist.csv, monitor_history.csv — durable watchlist + snapshots.
• Per-cycle ${WORKSPACE}/competitor-intel-[client]-[date].md reports.

Credentials / env

• Required: none to produce reports — backbone + diffing run keyless on local CSV state.
• Optional: SUPABASE_URL/SUPABASE_KEY or AIRTABLE_API_KEY (durable shared watchlist / history — strongly recommended); APIFY_API_TOKEN (Reddit/X/review depth); PHANTOMBUSTER_API_KEY + LinkedIn cookie (LinkedIn post monitoring); DATAFORSEO_LOGIN/ DATAFORSEO_PASSWORD or SERPER_API_KEY for per-cycle discovery searches (if set -> paid SERP; if not -> the agent's own web search, the default); SLACK_* / DISCORD_* / TELEGRAM_* (report delivery, else workspace file).

Notes & edge cases

• Stateful + scheduled: durable storage is what makes diffs meaningful — an in-flow-only run reports one cycle but can't detect cross-cycle change. The CSV store gives offline history.
• Respect the cadence to control proxy/scrape cost (reviews monthly, baseline quarterly).
• Negative-review trends and competitor-trouble signals are poaching triggers — surface them prominently, route to outreach only after the human checkpoint.
• Always gate recommended actions behind human approval (the two checkpoints).
• Apify degrade (when set): curl -s "https://api.apify.com/v2/acts/<actor>/run-sync-get-dataset-items?token=$APIFY_API_TOKEN" -d '{...}'.