end-to-end-news-signal

name	end-to-end-news-signal
version	1.0.0
description	Full pipeline from news signal discovery to outreach-ready campaign. Scans the web for recent news events — regulation changes, acquisitions, product launches, market shifts, crises, partnerships, earnings, and any other public event — qualifies affected companies against your ICP, identifies the connection between the news and your product, finds decision-makers, drafts personalized email sequences using the news as the hook, and packages for your outreach tool. Combines multi-source signal discovery with news-signal-outreach into a single zero-to-campaign workflow. No company list required.
tags	["lead-generation","outreach"]
graph	{"provides":["discovered-news-signals","icp-qualified-targets","connection-angles","contact-list","personalized-email-sequences","campaign-package"],"requires":["your-company-context","target-news-categories"],"connects_to":[{"skill":"cold-email-outreach","when":"User wants to launch the campaign via Smartlead or other outreach tool","passes":"contact-list, personalized-email-sequences"},{"skill":"linkedin-outreach","when":"User wants LinkedIn outreach alongside email","passes":"contact-list"}],"capabilities":["web-search","apify-search","contact-finding","email-drafting"]}

End-to-End News Signal

Full pipeline from news signal discovery to outreach-ready campaign in a single run. No company list required — this skill discovers relevant news events from scratch across multiple sources, extracts affected companies, qualifies them against your ICP, identifies the connection between the news and your product, finds the right people, drafts personalized emails using the news as the hook, and packages everything for your outreach tool.

This is the broadest end-to-end signal composite. While funding, hiring, and leadership change composites handle specific, predictable signal types, this one handles the long tail — regulation changes, acquisitions, product launches, market shifts, crises, partnerships, earnings events, and any other public event that creates a reason to reach out.

When to Auto-Load

Load this composite when:

• User says "find news I can use for outreach", "news-based prospecting", "end-to-end news signal"
• User says "scan for news in [industry] and reach out to affected companies"
• User wants to go from zero to outreach campaign based on recent news events
• User has no existing company list but wants to find companies affected by recent events
• User mentions a news category (regulation, acquisition, market shift) and wants outreach built around it
• User says "find companies affected by [event/trend] and reach out"

Do NOT load if:

• User already has a specific news item or URL → use news-signal-outreach instead
• User already has a company list and wants news angles → use news-signal-outreach (Mode B)
• The signal is specifically funding → use end-to-end-funding-signal
• The signal is specifically hiring → use end-to-end-hiring-signal
• The signal is specifically a leadership change → use end-to-end-leadership-change-signal

Architecture

This composite chains multi-source signal discovery into news-signal-outreach (qualification → connection angle → contacts → emails → handoff). The discovery layer's output feeds directly into the outreach pipeline.

┌──────────────────────────────────────────────────────────────────────────────────┐
│                        END-TO-END NEWS SIGNAL                                    │
│                                                                                  │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────┐  ┌────────┐  │
│  │ DISCOVER │─▶│ EXTRACT  │─▶│ QUALIFY  │─▶│  FIND    │─▶│DRAFT │─▶│ LAUNCH │  │
│  │ Signals  │  │ & Angle  │  │ & Rank   │  │ People   │  │Emails│  │        │  │
│  └──────────┘  └──────────┘  └──────────┘  └──────────┘  └──────┘  └────────┘  │
│       │              │             │              │            │          │       │
│  Multi-source   Parse news,   ICP filter +   Buyer        News-as-    Tool-     │
│  web scan       ID connection  prioritize    personas +    hook email  agnostic  │
│  (no input      angle          by strength   contact tool  sequences   export    │
│   list)                                                                          │
└──────────────────────────────────────────────────────────────────────────────────┘

---

Step 0: Configuration (One-Time Setup)

On first run, collect and store all configuration needed for the full pipeline. Skip on subsequent runs.

News Discovery Config

Question	Options	Stored As
What news categories are most relevant to your product?	regulation, acquisition, product launch, market expansion, cost cutting, partnership, crisis, earnings, industry trend, competitive move (multi-select)	target_news_categories
Any industry filters?	SaaS, AI, fintech, healthtech, cybersecurity, etc. (or "all")	target_industries
Any geography filters?	US, EMEA, APAC, etc. (or "all")	target_geographies
How far back should we search?	3 / 7 / 14 / 30 days	lookback_days
Any keywords or topics to prioritize?	Free text (e.g. "HIPAA", "AI regulation", "supply chain")	priority_keywords

ICP Definition

Question	Purpose	Stored As
What does your company do? (1-2 sentences)	Relevance matching	company_description
What problem do you solve?	Connection angle identification	pain_point
What company sizes do you sell to?	ICP filter	target_company_size
Any disqualifiers?	Hard no's	disqualifiers
What specific outcomes does your product deliver?	Relevance angle building	product_outcomes
Name 2-3 proof points (customers, metrics)	Email credibility	proof_points

Contact Finding Config

Question	Options	Stored As
Who are your buyers? (titles)	Free text	buyer_titles
Who are your champions? (titles)	Free text	champion_titles
How should we find contacts?	Apollo / LinkedIn Sales Nav / Clearbit / Web search / Manual	contact_tool
Do you have API access?	Yes (provide key) / No (use web search)	contact_api_access

Outreach Config

Question	Options	Stored As
Where do you want outreach sent?	Smartlead / Instantly / Outreach.io / Lemlist / Apollo / CSV export	outreach_tool
Email or multi-channel?	Email only / Email + LinkedIn	outreach_channels

Store config in: clients/<client-name>/config/end-to-end-news-signal.json or equivalent.

---

Step 1: Discover News Signals

Purpose: Scan multiple web sources for recent news events matching the target categories, industries, and keywords. No company list input needed.

Input Contract

target_news_categories: string[]    # From config
target_industries: string[]         # From config (optional)
target_geographies: string[]        # From config (optional)
lookback_days: integer              # From config (default: 7)
priority_keywords: string[]         # From config (optional)

Process

Run all source searches in parallel:

A) Web Search (WebSearch tool — free)

Run 6-10 varied queries combining news categories with industries:

• "[industry] [news_category] announced this week 2026"
• "new regulation [industry] 2026"
• "[industry] acquisition announced"
• "[industry] product launch 2026"
• "[industry] expansion layoffs restructuring"
• "[priority_keyword] news 2026"
• "[industry] company acquired OR partnered OR launched OR expanded"

For each result, extract: headline, event category, companies mentioned, date, source URL.

B) Twitter Search (twitter-scraper — ~$0.05-0.10)

python3 skills/twitter-scraper/scripts/search_twitter.py \
  --query "\"[priority_keyword]\" AND ([industry] OR [news_category_terms])" \
  --since <lookback-start> --until <today> --max-tweets 50 --output json

News often breaks on Twitter first. Founder and executive posts signal events before press coverage.

C) Hacker News (hacker-news-scraper — free)

python3 skills/hacker-news-scraper/scripts/search_hn.py \
  --query "[priority_keyword] OR [industry] [news_category]" \
  --days <lookback_days> --min-points 5 --output json

Tech and business news surfaces early on HN, especially regulation, launches, and acquisitions.

D) Reddit Search (reddit-scraper — ~$0.05-0.10)

python3 skills/reddit-scraper/scripts/search_reddit.py \
  --subreddit "[relevant_subreddits]" \
  --keywords "[priority_keywords],[news_category_terms]" \
  --days <lookback_days> --sort hot --output json

Industry-specific subreddits surface news with practitioner commentary that reveals impact.

Consolidation

1. Deduplicate across sources. Same news event from multiple sources = higher confidence.
2. Filter by target categories, industries, and geographies.
3. Group by event — multiple articles about the same event collapse into one signal with multiple sources.
4. Score each signal:
   • +3: Appears in multiple sources
   • +3: Category matches target_news_categories exactly
   • +2: Contains priority_keywords
   • +2: Industry matches target_industries
   • +1: Published within last 3 days
   • +1: High engagement (comments, shares, upvotes)
   • -1: Category outside target set
   • -2: Geography mismatch (if filter specified)
5. Rank by score descending.

Output Contract

discovered_signals: [
  {
    headline: string
    event_category: string            # "regulation", "acquisition", "expansion", etc.
    event_date: string
    full_summary: string              # 2-3 sentence summary
    source_urls: string[]
    sources: string[]                 # ["web", "twitter", "hn", "reddit"]
    companies_mentioned: [
      {
        name: string
        domain: string | null
        role_in_news: string          # "subject", "affected", "partner", "competitor"
        industry: string
      }
    ]
    people_mentioned: [
      {
        name: string
        title: string | null
        company: string
        role_in_news: string
      }
    ]
    affected_industries: string[]
    confidence: "high" | "medium"
    score: integer
    priority_keyword_matches: string[]
  }
]

Human Checkpoint

Found X news signals matching your criteria across Y sources:

| Rank | Headline | Category | Date | Companies | Sources | Score |
|------|----------|----------|------|-----------|---------|-------|
| 1    | FDA announces new telehealth regulation | Regulation | 2026-04-03 | 5 mentioned, ~50 affected | web, hn, reddit | 11 |
| 2    | Acme acquires BetaCo for $200M | Acquisition | 2026-04-01 | 2 mentioned | web, twitter | 8 |
| 3    | New AI safety framework released | Regulation | 2026-04-02 | 3 mentioned | web, hn | 7 |
| ...  | ...      | ...      | ...  | ...       | ...     | ...   |

Proceed to extraction and qualification? You can remove any signals before continuing.

---

Step 2: Extract Entities & Qualify Against ICP

Purpose: For each discovered signal, extract all affected companies (including those implied but not mentioned), then qualify against ICP. Combines the Parse & Extract and ICP Qualification steps from news-signal-outreach.

Input Contract

discovered_signals: [...]            # From Step 1 output
icp_criteria: {
  target_industries: string[]
  target_company_size: string
  target_geographies: string[]
  disqualifiers: string[]
}
your_company: {
  description: string
  pain_point: string
}

Process

For each discovered signal:

1. Extract directly mentioned companies and people from the signal summary and source content.

2. Expand to implied companies:

   • "New FDA regulation on telehealth" → search for telehealth companies, not just those in the article
   • "Major data breach at [company]" → the breached company AND their competitors
   • "Industry report shows X trend" → companies in that industry
   • Use web search to identify specific companies in affected spaces

3. Qualify each company against ICP:

   • Check industry against target_industries
   • Estimate company size (from news context or quick web search)
   • Check geography if relevant
   • Check against disqualifiers

4. Assign priority tiers:

   • Tier 1 (Act Today): Direct subject of the news + strong ICP fit + event within 7 days
   • Tier 2 (Act This Week): Directly affected by the news + ICP fit + event within 14 days
   • Tier 3 (Queue): Indirectly affected or thematic connection + ICP fit
   • Drop: No ICP fit or disqualified

Output Contract

qualified_targets: [
  {
    entity: {
      type: "company" | "person"
      name: string
      company: string
      domain: string | null
      role_in_news: string
      industry: string
      estimated_size: string | null
    }
    news: {
      headline: string
      event_category: string
      event_date: string
      full_summary: string
      source_urls: string[]
    }
    icp_assessment: {
      fit: "strong" | "moderate"
      industry_match: boolean
      size_match: boolean | "unknown"
      reasoning: string
    }
    priority_tier: "tier_1" | "tier_2" | "tier_3"
  }
]
dropped_targets: [
  { name: string, reason: string }
]

Human Checkpoint

## Extraction & Qualification Results

### Signal: "FDA announces new telehealth regulation"

#### Tier 1 — Act Today (X companies)
| Company | Role in News | Industry | Size | ICP Fit | Why |
|---------|-------------|----------|------|---------|-----|
| TeleHealth Co | Subject | Healthcare SaaS | ~200 | Strong | Core ICP, directly impacted |

#### Tier 2 — Act This Week (X companies)
| Company | Role in News | Industry | Size | ICP Fit | Why |
|---------|-------------|----------|------|---------|-----|
| MedTech Inc | Affected | HealthTech | ~500 | Moderate | Adjacent, needs compliance |

#### Tier 3 — Queue (X companies)
| Company | Role in News | Industry | Size | ICP Fit | Why |
|---------|-------------|----------|------|---------|-----|

#### Dropped (X)
| Company | Reason |
|---------|--------|

[Repeat for each signal]

Approve before we identify connection angles?

---

Step 3: Identify Connection Angles

Purpose: The critical thinking step. For each qualified target, determine the specific connection between the news event and your product. Why should they care about your product RIGHT NOW because of THIS news? Pure LLM reasoning.

Input Contract

qualified_targets: [...]              # From Step 2 output
your_company: {
  description: string
  pain_point: string
  product_outcomes: string[]
  proof_points: string[]
}

Process

For each qualified target, answer three questions:

Question 1: "Why does this news create urgency for our product?"

Map the news event category to a product relevance pattern:

Event Category	How It Creates Urgency	Example
Regulation change	They need to comply, your product helps them comply or adapt faster	"New data privacy law → they need [your compliance tool] before enforcement date"
Acquisition / Merger	Systems need integration, processes need standardization	"Acquired a company → need to unify [function your product handles]"
Market expansion	New market = new challenges, need tools that scale	"Expanding to EMEA → need [your product] for localized [function]"
Product launch	Scaling up means scaling operations	"Launching enterprise tier → need [your product] to handle enterprise [function]"
Competitive pressure	Competitor moved, they need to respond	"Competitor launched [X] → they need to level up [area your product addresses]"
Cost cutting / Layoffs	Do more with less, automation becomes essential	"Cut 15% of staff → need [your product] to maintain output with smaller team"
Crisis / Incident	Reactive buying — they need a solution NOW	"Data breach → urgently need [your security product]"
Partnership	New partner = new workflows	"Partnered with [company] → need [your product] to support the integration"
Earnings / Growth	Over-performing = scaling. Under-performing = efficiency	"Revenue grew 3x → [function your product handles] can't keep up manually"
Industry trend	Category awareness is high, they're evaluating	"Industry report says [trend] → they're likely evaluating solutions"

Question 2: "What's the specific angle?"

Craft a one-sentence connection:

"Because [news event], [company] now needs [specific outcome your product delivers]."

Question 3: "How strong is this connection?"

Strength	Criteria	Example
Direct	The news explicitly describes a problem your product solves	Layoff in your product's department → they need automation
Adjacent	The news implies a downstream need your product addresses	Market expansion → implies scaling → need for your tool
Thematic	The news is in the same category as your product's domain	Industry report about the trend you're in → awareness play

Framework Selection

Connection Strength	Framework	Why
Direct	Signal-Proof-Ask	The news IS the hook — reference it directly, show proof, ask
Adjacent	PAS	Problem (implied by the news) → Agitate → Solve
Thematic	AIDA	Attention (news) → Interest → Desire → Action

Output Contract

connection_angles: [
  {
    entity: { ... }
    news: { ... }
    icp_assessment: { ... }
    priority_tier: string
    connection: {
      urgency_reason: string
      specific_angle: string
      connection_strength: "direct" | "adjacent" | "thematic"
      timing_note: string
      recommended_framework: string
    }
  }
]

Human Checkpoint

## Connection Angles

### Direct Connections (X targets) — Act quickly
| Entity | News | Angle | Timing |
|--------|------|-------|--------|
| Acme Corp | Acquired BetaCo | "Need to unify CRM systems — [product] does this in 30 days" | This week |

### Adjacent Connections (X targets)
| Entity | News | Angle | Timing |
|--------|------|-------|--------|
| MedTech Inc | Expanding to EMEA | "Localized [function] becomes a requirement" | This month |

### Thematic Connections (X targets)
| Entity | News | Angle | Timing |
|--------|------|-------|--------|
| HealthCo | Industry report on [trend] | "Likely evaluating [category] solutions" | Flexible |

Approve these angles before we find contacts?

---

Step 4: Find Decision-Makers

Purpose: For each qualified target with a connection angle, find people matching buyer/champion personas, prioritized by relevance to the news event.

Input Contract

connection_angles: [...]              # From Step 3 output
buyer_titles: string[]                # From config
champion_titles: string[]             # From config
contact_tool: string                  # From config
max_contacts_per_company: integer     # Default: 3-5

Process

1. For each company, use the configured contact_tool to find people matching buyer and champion titles.

2. Prioritize contacts by news relevance:

   News Category	Prioritize These Contacts
   Regulation / Compliance	Legal, Compliance, Operations leadership
   Acquisition / Merger	COO, CTO, VP Operations, Integration leads
   Market expansion	VP Sales, VP Marketing, Country/Regional leads
   Cost cutting / Layoffs	COO, CFO, VP Operations
   Product launch	CTO, VP Product, VP Engineering
   Crisis / Incident	CISO, VP Engineering, CTO (security), CEO/COO (operational)
   General growth	Default to buyer_titles from config

3. For each contact, note their relevance to the news — are they directly affected, the decision-maker for the response, or the person who feels the pain?

4. Cap at max_contacts_per_company (3-5).

5. Deduplicate against contact-cache to avoid repeat outreach across runs.

Output Contract

contacts: [
  {
    person: {
      full_name: string
      first_name: string
      last_name: string
      title: string
      email: string | null
      linkedin_url: string | null
      role_type: "buyer" | "champion" | "user"
      news_relevance: string
    }
    company: {
      name: string
      domain: string
      priority_tier: string
    }
    connection: {
      specific_angle: string
      connection_strength: string
      urgency_reason: string
    }
    news: {
      headline: string
      event_category: string
      source_url: string | null
    }
  }
]
contacts_without_email: [...]         # Same structure, flagged for manual lookup

Human Checkpoint

## Contacts Found

### Acme Corp — "Acquired BetaCo" (Direct connection, Tier 1)
| Name | Title | Role | Why This Person | Email |
|------|-------|------|----------------|-------|
| Sarah Kim | COO | Buyer | Owns post-acquisition integration | ✓ |
| David Park | VP Operations | Champion | Manages unified workflows | ✓ |

### MedTech Inc — "Expanding to EMEA" (Adjacent connection, Tier 2)
| Name | Title | Role | Why This Person | Email |
|------|-------|------|----------------|-------|
| ... | ... | ... | ... | ... |

Total: X contacts across Y companies (Z without email)

Approve before we draft emails?

---

Step 5: Draft Personalized Emails

Purpose: Draft outreach where the news event is the hook, your product is the solution, and the email demonstrates you understand their specific situation. Pure LLM reasoning.

Input Contract

contacts: [...]                       # From Step 4 output
your_company: {
  description: string
  pain_point: string
  product_outcomes: string[]
  proof_points: string[]
}
sequence_config: {
  touches: integer                    # Default: 3
  timing: integer[]                   # Default varies by connection strength
  tone: string                        # Default: "casual-direct"
  cta: string                         # Default: "15-min call"
}

Process

1. Adjust sequence timing by connection strength:

   Strength	Timing	Rationale
   Direct	Day 1 / 3 / 7	Urgency is real — they're actively dealing with this
   Adjacent	Day 1 / 5 / 12	Standard — urgency is implied, not immediate
   Thematic	Day 1 / 7 / 14	Slower cadence — awareness, not crisis response

2. Build the email around the news, not the product. The news is the subject. Your product is the punchline.

   Element	Source	How to Use
   News hook	Step 1 headline	Open with what happened — show you're informed
   Impact on them	Step 3 urgency_reason	Explain what this means for their specific role
   Your angle	Step 3 specific_angle	Connect the dots to your product naturally
   Proof	Config proof_points	Show a peer who faced a similar situation
   CTA	Config	Low-friction ask

3. Email structure by connection strength:

   Direct connection (Signal-Proof-Ask):

   Hook: Reference the specific news event
   Impact: What this means for them (1 sentence)
   Proof: A peer who faced the same situation and used your product
   Ask: Soft CTA
   

   Adjacent connection (PAS):

   Problem: The downstream challenge the news creates
   Agitate: What happens if they don't address it (1 sentence)
   Solve: How your product helps, with a proof point
   Ask: Soft CTA
   

   Thematic connection (AIDA):

   Attention: Reference the news/trend
   Interest: How it relates to their company specifically
   Desire: What your product does in this context
   Action: Soft CTA
   

4. Generate emails per email-drafting skill rules:

Output Contract

email_sequences: [
  {
    contact: { full_name, email, title, company_name, role_type, news_relevance }
    news_context: { headline, event_category, source_url }
    connection: { specific_angle, connection_strength }
    sequence: [
      {
        touch_number: integer
        send_day: integer
        subject: string
        body: string
        framework: string
        word_count: integer
      }
    ]
  }
]

Human Checkpoint

Present 3-5 sample sequences (one per connection strength / tier combination):

## Sample Emails for Review

### Direct Connection: Sarah Kim, COO @ Acme Corp
News: Acme acquired BetaCo | Angle: CRM unification | Framework: Signal-Proof-Ask

**Touch 1 — Day 1**
Subject: Unifying Acme + BetaCo systems
> Hi Sarah — saw the BetaCo acquisition. Congrats. The integration
> sprint typically surfaces a CRM unification challenge fast —
> two systems, overlapping data, different workflows.
>
> [Peer company] faced the same thing after their acquisition last year.
> [Product] had both systems unified in 30 days. Happy to share how.
>
> Worth a 15-minute call?

**Touch 2 — Day 3**
> [New angle — data migration complexity, with a specific metric]

**Touch 3 — Day 7**
> [Breakup with offer to share the integration playbook]

---

### Adjacent Connection: Dr. Lee, VP Product @ MedTech Inc
News: EMEA expansion | Angle: Localization needs | Framework: PAS

**Touch 1 — Day 1**
Subject: EMEA expansion + [function] localization
> [full email]

---

Approve these samples? I'll generate the rest in the same style.
Iterate? Tell me what to change (tone, length, angle, CTA).

---

Step 6: Launch Campaign

Purpose: Package contacts + email sequences for the configured outreach tool.

Input Contract

email_sequences: [...]                # From Step 5 output
outreach_tool: string                 # From config
outreach_channels: string             # From config

Process

Tool	Action
Smartlead	Chain to cold-email-outreach Smartlead MCP automation
Instantly	Generate Instantly-format CSV
Outreach.io	Generate Outreach-compatible CSV
Lemlist	Generate Lemlist-format CSV
Apollo	Generate Apollo sequence import CSV
CSV export	Generate generic CSV with all fields

If outreach_channels includes LinkedIn:

• Chain to linkedin-outreach for LinkedIn message sequences
• Output CSV for LinkedIn automation tool

Output Contract

campaign_package: {
  tool: string
  file_path: string
  contact_count: integer
  sequence_touches: integer
  estimated_send_days: integer
  signals_used: integer
  connection_breakdown: {
    direct: integer
    adjacent: integer
    thematic: integer
  }
  next_action: string
}

Human Checkpoint

## Campaign Ready

Tool: [configured tool]
Signal type: News-triggered
Signals used: X news events
Contacts: N people across M companies
Connection strengths: X direct, Y adjacent, Z thematic
Tier breakdown: A Tier 1, B Tier 2, C Tier 3
Sequence: 3 touches (timing varies by connection strength)
File: output/{campaign-name}-{date}.csv

Ready to launch? (Final gate before emails are sent or files are created)

---

Execution Summary

Step	What	Tool	Checkpoint	Time
0	Config	None	First run only	5 min (once)
1	Discover news signals	WebSearch + Twitter + HN + Reddit	Review discovered signals	3-5 min
2	Extract & qualify	Web search (expansion) + LLM reasoning	Approve qualified targets	3-5 min
3	Connection angles	LLM reasoning	Approve angles + strength ratings	3-5 min
4	Find people	Configurable (Apollo, LinkedIn, etc.)	Approve contact list	2-3 min
5	Draft emails	LLM reasoning	Review samples, iterate	5-10 min
6	Launch	Configurable (Smartlead, CSV, etc.)	Final approval	1 min

Total: ~20-30 minutes from "find me news I can use for outreach" to outreach-ready campaign.

Cost

Component	Cost
Web Search	Free
Hacker News (Algolia)	Free
Twitter scraper (Apify)	~$0.05-0.10
Reddit scraper (Apify)	~$0.05-0.10
Contact finding (varies by tool)	$0-0.50

Typical run: $0.10-0.70 total depending on contact tool.

Tips

• Direct connections are rare but powerful. Most news creates adjacent or thematic connections. When you find a direct one, prioritize it — these convert at 2-3x the rate.
• Speed matters for direct connections. The first vendor to reference a relevant news event looks informed. The fifth looks like they're running the same playbook.
• Don't force weak connections. If you can't articulate the angle in one sentence, the connection is too weak. Drop it.
• News about competitors is gold. If a competitor raises funding, gets acquired, has a breach, or launches a product — their customers are suddenly open to conversations.
• Negative news requires a 48-72 hour cooling period. Reaching out the day of a layoff or breach is predatory. Wait, then lead with empathy.
• Industry reports and trend pieces make great thematic triggers. "The Gartner report on [category] just dropped — here's what it means for [company]."
• Run weekly — fresh news keeps outreach relevant and timely.
• 3-5 contacts per company avoids carpet-bombing.
• Track contacts in contact-cache to avoid duplicate outreach across weekly runs.
• Route embedded signals to specialist composites. News often contains funding, hiring, or leadership signals. Route those to the dedicated composite for better outreach.
• Multi-source appearances are the strongest signal — a story on TechCrunch + HN + Twitter is higher quality than a single-source mention.

Example Prompt

"Scan for news in cybersecurity and healthtech from the past week. We sell compliance automation for mid-market SaaS companies. Our buyers are VP Engineering and CISOs. Focus on regulation changes, data breaches, and acquisition activity. Find contacts via web search and export to CSV."

The agent should run the full pipeline: discover signals → extract & qualify → connection angles → find contacts → draft emails → export CSV.