Business operating system for Claude Code — 57 skills, 21 agents, smart daemon. Unified inbox (WhatsApp/Email/Slack/Telegram), autonomous PR merge, full-AWS monitoring, revenue (Stripe+RevenueCat), e-commerce (Shopify), marketing (Klaviyo/Meta/GA4), voice (Bland/ElevenLabs), APM (Datadog/NewRelic/OTEL), YOLO mode.
Shell agent-teamsai-agentai-automationanthropicautonomous-agentawsbusiness-operationsclaudeclaude-codeclaude-code-plugindevopsecommerceinbox-zeroklaviyomarketingmcpmorning-briefingpr-automationshopifyterminal
feat(competitor-intel): self-discovering LLM-driven analyzer with weekly delta detection (#237)
Aurora Capitalb31e8e3e2869
diff --git a/claude-ops/scripts/ops-cron-competitor-intel.sh b/claude-ops/scripts/ops-cron-competitor-intel.shindex a0a7ec54..33ce5f25 100755--- a/claude-ops/scripts/ops-cron-competitor-intel.sh+++ b/claude-ops/scripts/ops-cron-competitor-intel.sh@@ -1,34 +1,50 @@#!/usr/bin/env bash-# ops-cron-competitor-intel.sh — Weekly Competitor Intel cron job-# Searches competitors and own brand reviews, reports to Telegram-# Configure via /ops:setup (writes to preferences.json) OR env vars:-# COMPETITOR_A_QUERY, COMPETITOR_B_QUERY, BRAND_QUERY, REPORT_TIMEZONE-# Env vars override preferences.json when both are set.+# ops-cron-competitor-intel.sh — Weekly self-discovering competitor intel+#+# Pipeline (per brand seed):+# 1. Tavily discovery → top competitors to {brand} in {category}+# 2. State merge → diff vs competitor_state.json, flag NEW entrants+# 3. Tavily news pass → recent moves per top-5 competitor (last 7d)+# 4. Tavily brand pass → own-brand reviews / mentions+# 5. LLM synthesis → claude_invoke (Sonnet) one-page strategic delta+# 6. Telegram digest + persist new state for next week's delta+#+# Config (preferences.json → .competitor_intel):+# brand_name — e.g. "My-Project" (REQUIRED)+# category — e.g. "AI health coaching apps" (REQUIRED)+# max_competitors — int, default 5+# report_timezone — IANA TZ, default UTC+#+# Env overrides: BRAND_NAME, BRAND_CATEGORY, MAX_COMPETITORS, REPORT_TIMEZONE+#set -euo pipefail+PLUGIN_ROOT="${CLAUDE_PLUGIN_ROOT:-$HOME/.claude/plugins/cache/ops-marketplace/ops/$(ls -1 "$HOME/.claude/plugins/cache/ops-marketplace/ops/" 2>/dev/null | sort -V | tail -1)}"DATA_DIR="${OPS_DATA_DIR:-$HOME/.claude/plugins/data/ops-ops-marketplace}"LOG_DIR="$DATA_DIR/logs"LOG="$LOG_DIR/competitor-intel.log"PREFS_PATH="$DATA_DIR/preferences.json"+STATE_PATH="$DATA_DIR/competitor_state.json"mkdir -p "$LOG_DIR"log() { printf '%s [competitor-intel] %s\n' "$(date -u +%Y-%m-%dT%H:%M:%SZ)" "$1" | tee -a "$LOG"; }-# ── Resolve query config from preferences.json (env vars take precedence) ─+# ── Resolve config ────────────────────────────────────────────────────────pref_get() {local key="$1"[[ -f "$PREFS_PATH" ]] || { echo ""; return; }jq -r --arg k "$key" '.competitor_intel[$k] // ""' "$PREFS_PATH" 2>/dev/null || echo ""}-COMPETITOR_A_QUERY="${COMPETITOR_A_QUERY:-$(pref_get competitor_a_query)}"-COMPETITOR_B_QUERY="${COMPETITOR_B_QUERY:-$(pref_get competitor_b_query)}"-BRAND_QUERY="${BRAND_QUERY:-$(pref_get brand_query)}"+BRAND_NAME="${BRAND_NAME:-$(pref_get brand_name)}"+BRAND_CATEGORY="${BRAND_CATEGORY:-$(pref_get category)}"+MAX_COMPETITORS="${MAX_COMPETITORS:-$(pref_get max_competitors)}"REPORT_TIMEZONE="${REPORT_TIMEZONE:-$(pref_get report_timezone)}"+MAX_COMPETITORS="${MAX_COMPETITORS:-5}"+REPORT_TIMEZONE="${REPORT_TIMEZONE:-UTC}"-# ── Refuse to run with placeholder queries ────────────────────────────────-if [[ -z "$COMPETITOR_A_QUERY" && -z "$COMPETITOR_B_QUERY" && -z "$BRAND_QUERY" ]]; then- log "SKIP: no competitor/brand queries configured. Run /ops:setup to configure."+if [[ -z "$BRAND_NAME" || -z "$BRAND_CATEGORY" ]]; then+ log "SKIP: brand_name + category not configured in preferences.json. Run /ops:setup."log "HEARTBEAT_OK"exit 0fi@@ -45,79 +61,166 @@ if [[ -z "$TAVILY_KEY" ]]; thenTAVILY_KEY=$(doppler secrets get TAVILY_API_KEY --plain 2>/dev/null || true)fi+if [[ -z "$TAVILY_KEY" ]]; then+ log "SKIP: TAVILY_API_KEY not available — can't run intelligent search."+ log "HEARTBEAT_OK"+ exit 0+fi+YEAR=$(date +%Y)-FINDINGS=()--# ── Search function via Tavily (or fallback curl) ─────────────────────────-search_web() {- local query="$1"- local label="$2"- local result=""-- if [[ -n "$TAVILY_KEY" ]]; then- result=$(curl -s -X POST "https://api.tavily.com/search" \- -H "Content-Type: application/json" \- -H "Authorization: Bearer ${TAVILY_KEY}" \- -d "{\"query\": \"$query\", \"max_results\": 3, \"search_depth\": \"basic\"}" \- 2>/dev/null | python3 -c "-import json, sys-data = json.load(sys.stdin)-results = data.get('results', [])-if not results:- print('No results found')-else:- snippets = [r.get('title','') + ': ' + r.get('content','')[:120] for r in results[:2]]- print(' | '.join(snippets))-" 2>/dev/null || echo "search unavailable")- else- # Fallback: SerpAPI-compatible curl search via DuckDuckGo HTML (best-effort)- result=$(curl -s -A "Mozilla/5.0" \- "https://html.duckduckgo.com/html/?q=$(python3 -c "import urllib.parse,sys; print(urllib.parse.quote('$query'))")" \- 2>/dev/null | python3 -c "-import sys, re-html = sys.stdin.read()-titles = re.findall(r'class=\"result__title\"[^>]*>.*?<a[^>]*>(.*?)</a>', html, re.DOTALL)-clean = [re.sub('<[^>]+>','',t).strip() for t in titles[:3]]-print(' | '.join(clean[:2]) if clean else 'No results')-" 2>/dev/null || echo "search unavailable")- fi- log "Search [$label]: ${result:0:100}..."- echo "$result"+# ── Tavily helpers ────────────────────────────────────────────────────────+tavily_search() {+ # $1=query, $2=depth(basic|advanced), $3=topic(general|news), $4=days+ local query="$1" depth="${2:-basic}" topic="${3:-general}" days="${4:-30}"+ local payload+ payload=$(jq -n --arg q "$query" --arg d "$depth" --arg t "$topic" --argjson days "$days" \+ '{query:$q, search_depth:$d, topic:$t, max_results:5, days:$days}')++ curl -s -X POST "https://api.tavily.com/search" \+ -H "Content-Type: application/json" \+ -H "Authorization: Bearer ${TAVILY_KEY}" \+ -d "$payload" 2>/dev/null || echo '{"results":[]}'}-# ── Run searches ──────────────────────────────────────────────────────────-log "Running weekly competitor intelligence searches"+# ── Stage 1: Discovery — surface current competitor landscape ─────────────+log "Stage 1: discovering competitors to '$BRAND_NAME' in category '$BRAND_CATEGORY'"+DISCOVERY_QUERY="Top competitors to $BRAND_NAME in $BRAND_CATEGORY market $YEAR — list company names"+DISCOVERY_JSON=$(tavily_search "$DISCOVERY_QUERY" "advanced" "general" 90)+DISCOVERY_TEXT=$(echo "$DISCOVERY_JSON" | jq -r '.results[]? | "[\(.title // "?")] \((.content // "")[:400])"' 2>/dev/null | head -c 6000)++# ── Stage 2: Load prior state ─────────────────────────────────────────────+KNOWN_COMPETITORS="[]"+if [[ -f "$STATE_PATH" ]]; then+ KNOWN_COMPETITORS=$(jq -r --arg b "$BRAND_NAME" '.[$b].competitors // []' "$STATE_PATH" 2>/dev/null || echo "[]")+fi+log "Stage 2: $(echo "$KNOWN_COMPETITORS" | jq 'length // 0' 2>/dev/null || echo 0) competitors in prior state"++# ── Stage 3: Brand mentions (own reviews/news) ────────────────────────────+log "Stage 3: scanning brand mentions for '$BRAND_NAME'"+BRAND_NEWS_JSON=$(tavily_search "$BRAND_NAME reviews OR mentions OR launch OR funding" "basic" "news" 7)+BRAND_TEXT=$(echo "$BRAND_NEWS_JSON" | jq -r '.results[]? | "[\(.title // "?")] \((.content // "")[:300])"' 2>/dev/null | head -c 4000)++# ── Stage 4: Recent moves per known competitor (cap at MAX_COMPETITORS) ───+COMPETITOR_MOVES=""+if [[ "$KNOWN_COMPETITORS" != "[]" && "$KNOWN_COMPETITORS" != "" ]]; then+ log "Stage 4: scanning recent moves for known competitors"+ while IFS= read -r comp; do+ [[ -z "$comp" ]] && continue+ NEWS_JSON=$(tavily_search "$comp pricing OR launch OR funding OR product $YEAR" "basic" "news" 7)+ SNIP=$(echo "$NEWS_JSON" | jq -r '.results[]? | " • \(.title // "?"): \((.content // "")[:200])"' 2>/dev/null | head -3)+ [[ -n "$SNIP" ]] && COMPETITOR_MOVES+=$'\n## '"$comp"$'\n'"$SNIP"$'\n'+ done < <(echo "$KNOWN_COMPETITORS" | jq -r '.[]?' 2>/dev/null | head -n "$MAX_COMPETITORS")+fi++# ── Stage 5: LLM synthesis — strategic delta report ───────────────────────+log "Stage 5: LLM synthesis (Sonnet)"++NEW_STATE_COMPETITORS="$KNOWN_COMPETITORS"+REPORT=""++# shellcheck disable=SC1091+if [[ -f "$PLUGIN_ROOT/scripts/lib/claude-invoke.sh" ]]; then+ . "$PLUGIN_ROOT/scripts/lib/claude-invoke.sh"++ PROMPT_FILE=$(mktemp)+ cat > "$PROMPT_FILE" <<EOF+You are a strategic competitor intelligence analyst.++Brand: $BRAND_NAME+Category: $BRAND_CATEGORY+Date: $(TZ="$REPORT_TIMEZONE" date '+%Y-%m-%d')+Previously known competitors: $(echo "$KNOWN_COMPETITORS" | jq -c '.' 2>/dev/null || echo "[]")++## Raw Tavily research++### Discovery pass (current landscape):+$DISCOVERY_TEXT-COMPETITOR_A_RESULT=$([[ -n "$COMPETITOR_A_QUERY" ]] && search_web "$COMPETITOR_A_QUERY" "Competitor A" || echo "(not configured)")-COMPETITOR_B_RESULT=$([[ -n "$COMPETITOR_B_QUERY" ]] && search_web "$COMPETITOR_B_QUERY" "Competitor B" || echo "(not configured)")-BRAND_RESULT=$([[ -n "$BRAND_QUERY" ]] && search_web "$BRAND_QUERY" "Brand mentions" || echo "(not configured)")+### Brand mentions (last 7 days, $BRAND_NAME):+$BRAND_TEXT-# ── Build report ──────────────────────────────────────────────────────────-DATE_LABEL=$(TZ="${REPORT_TIMEZONE:-UTC}" date "+%a %d %b %Y")-REPORT="*Weekly Competitor Intel* ($DATE_LABEL)+### Recent competitor moves (last 7 days):+$COMPETITOR_MOVES-*Competitor A:*-$COMPETITOR_A_RESULT+## Your tasks-*Competitor B:*-$COMPETITOR_B_RESULT+1. Extract the canonical competitor list from the Discovery pass — return ONLY actual product/company names, not generic categories. Cap at $MAX_COMPETITORS.+2. Identify NEW entrants (in current list but not in "Previously known competitors").+3. Identify DROPPED competitors (in prior list but absent from current landscape).+4. Synthesize a one-page strategic delta. Skip filler. Strategic, not generic.-*Brand Mentions:*-$BRAND_RESULT"+Output format (STRICT — must be parseable):-log "Report built — sending to Telegram"+\`\`\`json+{"competitors": ["name1","name2"]}+\`\`\`++---REPORT---+*$BRAND_NAME Competitor Intel — $(TZ="$REPORT_TIMEZONE" date '+%a %d %b %Y')*++*NEW entrants:*+...++*Competitor moves this week:*+...++*Brand signal:*+...++*Threats & opportunities:*+...+EOF++ SYNTHESIS=$(claude_invoke --model claude-sonnet-4-6 --no-session-persistence -p < "$PROMPT_FILE" 2>>"$LOG" || echo "")+ rm -f "$PROMPT_FILE"++ if [[ -n "$SYNTHESIS" ]]; then+ EXTRACTED=$(echo "$SYNTHESIS" | sed -n '/```json/,/```/p' | sed '1d;$d' | jq -r '.competitors // []' 2>/dev/null || echo "")+ [[ -n "$EXTRACTED" && "$EXTRACTED" != "null" ]] && NEW_STATE_COMPETITORS="$EXTRACTED"+ REPORT=$(echo "$SYNTHESIS" | awk '/---REPORT---/{flag=1; next} flag' | sed '/^```/,/^```$/d')+ fi+else+ log "WARN: claude-invoke.sh not found at $PLUGIN_ROOT — using Tavily-only fallback"+fi++# Fallback report if LLM unavailable or returned empty+if [[ -z "$REPORT" ]]; then+ REPORT="*Weekly Competitor Intel — $(TZ="$REPORT_TIMEZONE" date '+%a %d %b %Y')*++*Discovery (raw):*+${DISCOVERY_TEXT:0:1200}++*Brand mentions (raw):*+${BRAND_TEXT:0:1200}"+fi++# ── Persist state ─────────────────────────────────────────────────────────+TMP_STATE=$(mktemp)+if [[ -f "$STATE_PATH" ]]; then+ jq --arg b "$BRAND_NAME" --argjson c "$NEW_STATE_COMPETITORS" --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \+ '.[$b] = {competitors: $c, last_run: $ts}' "$STATE_PATH" > "$TMP_STATE" 2>/dev/null \+ || jq -n --arg b "$BRAND_NAME" --argjson c "$NEW_STATE_COMPETITORS" --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \+ '{($b): {competitors: $c, last_run: $ts}}' > "$TMP_STATE"+else+ jq -n --arg b "$BRAND_NAME" --argjson c "$NEW_STATE_COMPETITORS" --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \+ '{($b): {competitors: $c, last_run: $ts}}' > "$TMP_STATE"+fi+mv "$TMP_STATE" "$STATE_PATH"+log "State persisted: $(echo "$NEW_STATE_COMPETITORS" | jq 'length // 0' 2>/dev/null || echo 0) competitors tracked for $BRAND_NAME"-# ── Send to Telegram ──────────────────────────────────────────────────────-if [[ -n "$TELEGRAM_TOKEN" ]]; then+# ── Stage 6: Send to Telegram ─────────────────────────────────────────────+log "Stage 6: dispatching Telegram digest"+if [[ -n "$TELEGRAM_TOKEN" && -n "$TELEGRAM_CHAT" ]]; thencurl -s -X POST "https://api.telegram.org/bot${TELEGRAM_TOKEN}/sendMessage" \-H "Content-Type: application/json" \- -d "{\"chat_id\": \"$TELEGRAM_CHAT\", \"text\": $(echo "$REPORT" | python3 -c 'import json,sys; print(json.dumps(sys.stdin.read()))'), \"parse_mode\": \"Markdown\"}" \+ -d "$(jq -n --arg chat "$TELEGRAM_CHAT" --arg text "$REPORT" \+ '{chat_id: $chat, text: $text, parse_mode: "Markdown"}')" \>> "$LOG" 2>&1log "Competitor intel sent to Telegram chat=$TELEGRAM_CHAT"else- log "WARN: TELEGRAM_BOT_TOKEN not set — printing report to stdout"- echo "$REPORT"+ log "WARN: TELEGRAM creds not set — printing report to stdout"+ printf '%s\n' "$REPORT"filog "HEARTBEAT_OK"diff --git a/claude-ops/skills/setup/SKILL.md b/claude-ops/skills/setup/SKILL.mdindex 6b2da2da..03de2ecb 100644--- a/claude-ops/skills/setup/SKILL.md+++ b/claude-ops/skills/setup/SKILL.md@@ -1040,40 +1040,39 @@ Write daemon services config to `$DATA_DIR/daemon-services.json` — merge with- `store-health`: `{ "enabled": true, "command": "${CLAUDE_PLUGIN_ROOT}/scripts/ops-cron-store-health.sh", "cron": "0 9 * * *" }` — daily 9am, only if ecom configured- `competitor-intel`: `{ "enabled": <bool>, "command": "${CLAUDE_PLUGIN_ROOT}/scripts/ops-cron-competitor-intel.sh", "cron": "0 10 * * 1" }` — weekly Monday 10am. `enabled` is `true` only when queries were configured in step 5b-i; otherwise `false` so the cron doesn't post placeholder garbage to Telegram-#### Step 5b-i — Competitor intel queries (gate before enabling the service)+#### Step 5b-i — Competitor intel (self-discovering, gate before enabling)Per Rule 3, never silently skip a service. Before deciding `competitor-intel.enabled`, ask the user explicitly:```AskUserQuestion({- question: "Configure competitor & brand intel queries? (Weekly search → Telegram)",+ question: "Configure competitor intel? (Weekly: Tavily auto-discovers competitors → Sonnet synthesizes strategic delta → Telegram)",header: "Competitor intel",options: [- { label: "Configure now", description: "Enter competitor A/B + own brand search queries — runs every Mon 10am" },+ { label: "Configure now", description: "Provide brand name + category. System auto-discovers competitors every week and detects deltas." },{ label: "Skip — disable", description: "Leave competitor-intel disabled. Re-run /ops:setup later to configure." }]})```-If the user chose **Skip — disable**: set `competitor-intel.enabled = false` in `daemon-services.json`. Do not prompt further.+If **Skip — disable**: set `competitor-intel.enabled = false` in `daemon-services.json`. Stop.-If the user chose **Configure now**: collect four values (one prompt per value — they are free-text, not multiple choice):+If **Configure now**: collect two free-text values plus one optional:-1. `competitor_a_query` — e.g. `"acme inc reviews"` or `"competitor-a $YEAR pricing"`-2. `competitor_b_query` — e.g. `"competitor-b new product launches"`-3. `brand_query` — e.g. `"yourbrand reviews"` (the user's own brand for reputation monitoring)-4. `report_timezone` — IANA TZ, e.g. `"Europe/Amsterdam"`. Default: `"UTC"`+1. `brand_name` — e.g. `"My-Project"`, `"Example-Project"`. The product/company being tracked.+2. `category` — e.g. `"AI health coaching apps"`, `"NL virtual real estate staging"`. The market segment Tavily searches in.+3. `report_timezone` — IANA TZ. Default: system TZ from previous setup steps, or `"UTC"`.-Persist to `$PREFS_PATH` under a top-level `competitor_intel` object:+The system handles competitor discovery, news scanning, and synthesis automatically — no hardcoded competitor list to maintain. Tavily API key (`TAVILY_API_KEY` in env or Doppler) is required; if missing, the cron logs SKIP and exits cleanly.++Persist to `$PREFS_PATH`:```bash-jq --arg a "$A" --arg b "$B" --arg brand "$BRAND" --arg tz "$TZ" \- '.competitor_intel = {competitor_a_query: $a, competitor_b_query: $b, brand_query: $brand, report_timezone: $tz}' \+jq --arg brand "$BRAND" --arg cat "$CATEGORY" --arg tz "$TZ" \+ '.competitor_intel = {brand_name: $brand, category: $cat, max_competitors: 5, report_timezone: $tz}' \"$PREFS_PATH" > "$PREFS_PATH.tmp" && mv "$PREFS_PATH.tmp" "$PREFS_PATH"```-The cron script (`ops-cron-competitor-intel.sh`) reads from `preferences.json` automatically — no env-var wiring needed in the daemon plist. Env vars still override `preferences.json` when set, for power users.-Now set `competitor-intel.enabled = true` in `daemon-services.json`.- `message-listener`: `{ "enabled": true, "command": "${CLAUDE_PLUGIN_ROOT}/scripts/ops-message-listener.sh" }` — only if WhatsApp or Telegram configured