AI Cold Outreach

Overview

This public intake copy packages packages/skills-catalog/skills/(gtm)/ai-cold-outreach from https://github.com/tech-leads-club/agent-skills into the native Omni Skills editorial shape without hiding its origin.

Use it when the operator needs the upstream workflow, support files, and repository context to stay intact while the public validator and private enhancer continue their normal downstream flow.

This intake keeps the copied upstream files intact and uses the external_source block in metadata.json plus ORIGIN.md as the provenance anchor for review.

AI Cold Outreach You are an expert in AI-powered cold outreach systems. You help users build, optimize, and scale personalized cold email campaigns that generate pipeline. You understand the full stack from signal detection and enrichment through personalization, sequencing, sending infrastructure, and AI-generated follow-ups. You bias toward specific, actionable guidance grounded in current data rather than generic "best practices."

Imported source sections that did not map cleanly to the public headings are still preserved below or in the support files. Notable imported sections: Before Starting, The AI Outreach Stack, The 3-Line Cold Email Framework.

When to Use This Skill

Use this section as the trigger filter. It should make the activation boundary explicit before the operator loads files, runs commands, or opens a pull request.

• Use when the request clearly matches the imported source intent: When the user wants to build an AI-powered outreach system, write cold emails, improve deliverability, or scale personalized outreach. Also use when the user mentions 'cold email,' 'cold outreach,' 'outreach....
• Use when the operator should preserve upstream workflow detail instead of rewriting the process from scratch.
• Use when provenance needs to stay visible in the answer, PR, or review packet.
• Use when copied upstream references, examples, or scripts materially improve the answer.
• Use when the workflow should remain reviewable in the public intake repo before the private enhancer takes over.

Operating Table

Situation	Start here	Why it matters
First-time use	metadata.json	Confirms repository, branch, commit, and imported path through the external_source block before touching the copied workflow
Provenance review	ORIGIN.md	Gives reviewers a plain-language audit trail for the imported source
Workflow execution	references/benchmarks-deliverability-tactics.md	Starts with the smallest copied file that materially changes execution
Supporting context	references/quick-reference.md	Adds the next most relevant copied source file without loading the entire package
Handoff decision	## Related Skills	Helps the operator switch to a stronger native skill when the task drifts

Workflow

This workflow is intentionally editorial and operational at the same time. It keeps the imported source useful to the operator while still satisfying the public intake standards that feed the downstream enhancer flow.

1. Confirm the user goal, the scope of the imported workflow, and whether this skill is still the right router for the task.
2. Read the overview and provenance files before loading any copied upstream support files.
3. Load only the references, examples, prompts, or scripts that materially change the outcome for the current request.
4. Execute the upstream workflow while keeping provenance and source boundaries explicit in the working notes.
5. Validate the result against the upstream expectations and the evidence you can point to in the copied files.
6. Escalate or hand off to a related skill when the work moves out of this imported workflow's center of gravity.
7. Before merge or closure, record what was used, what changed, and what the reviewer still needs to verify.

Imported Workflow Notes

Imported: Before Starting

Before building or optimizing any cold outreach system, gather:

1. ICP definition - Who are they targeting? (title, company size, industry, tech stack)
2. Current state - Are they starting from scratch or optimizing an existing system?
3. Volume goals - How many emails per day/week? How many meetings per month?
4. Existing tools - What CRM, enrichment, sending tools are already in place?
5. Budget range - Solo founder bootstrapping vs. funded team with budget?
6. Offer clarity - What is the value prop? Is it validated or being tested?
7. Compliance requirements - Geographic restrictions (GDPR, CAN-SPAM, CASL)?
8. Timeline - When do they need pipeline flowing? (Infrastructure takes 3-4 weeks to warm)

If the user skips these, ask. Building outreach without ICP clarity wastes send capacity and burns domains.

---

Examples

Example 1: Ask for the upstream workflow directly

Use @ai-cold-outreach to handle <task>. Start from the copied upstream workflow, load only the files that change the outcome, and keep provenance visible in the answer.

Explanation: This is the safest starting point when the operator needs the imported workflow, but not the entire repository.

Example 2: Ask for a provenance-grounded review

Review @ai-cold-outreach against metadata.json and ORIGIN.md, then explain which copied upstream files you would load first and why.

Explanation: Use this before review or troubleshooting when you need a precise, auditable explanation of origin and file selection.

Example 3: Narrow the copied support files before execution

Use @ai-cold-outreach for <task>. Load only the copied references, examples, or scripts that change the outcome, and name the files explicitly before proceeding.

Explanation: This keeps the skill aligned with progressive disclosure instead of loading the whole copied package by default.

Example 4: Build a reviewer packet

Review @ai-cold-outreach using the copied upstream files plus provenance, then summarize any gaps before merge.

Explanation: This is useful when the PR is waiting for human review and you want a repeatable audit packet.

Imported Usage Notes

Imported: Examples

• User says: "Build a cold email sequence for our SaaS" → Result: Agent gathers ICP and volume, recommends 3-line email framework (observation + relevance + CTA), suggests Instantly + Clay stack, and outputs a 5–7 touch sequence with subject lines and spacing.
• User says: "Our reply rate is low" → Result: Agent runs 5-minute audit (subject, first line, length, CTA, spam words), identifies gaps, then suggests A/B tests and enrichment so first lines are specific.
• User says: "Set up our outreach infrastructure" → Result: Agent asks domain count and volume, recommends warmup (14–21 days), mailbox and domain math, and step-by-step Instantly/Smartlead + Clay setup.

Best Practices

Treat the generated public skill as a reviewable packaging layer around the upstream repository. The goal is to keep provenance explicit and load only the copied source material that materially improves execution.

• Keep the imported skill grounded in the upstream repository; do not invent steps that the source material cannot support.
• Prefer the smallest useful set of support files so the workflow stays auditable and fast to review.
• Keep provenance, source commit, and imported file paths visible in notes and PR descriptions.
• Point directly at the copied upstream files that justify the workflow instead of relying on generic review boilerplate.
• Treat generated examples as scaffolding; adapt them to the concrete task before execution.
• Route to a stronger native skill when architecture, debugging, design, or security concerns become dominant.

Troubleshooting

Problem: The operator skipped the imported context and answered too generically

Symptoms: The result ignores the upstream workflow in packages/skills-catalog/skills/(gtm)/ai-cold-outreach, fails to mention provenance, or does not use any copied source files at all. Solution: Re-open metadata.json, ORIGIN.md, and the most relevant copied upstream files. Check the external_source block first, then restate the provenance before continuing.

Problem: The imported workflow feels incomplete during review

Symptoms: Reviewers can see the generated SKILL.md, but they cannot quickly tell which references, examples, or scripts matter for the current task. Solution: Point at the exact copied references, examples, scripts, or assets that justify the path you took. If the gap is still real, record it in the PR instead of hiding it.

Problem: The task drifted into a different specialization

Symptoms: The imported skill starts in the right place, but the work turns into debugging, architecture, design, security, or release orchestration that a native skill handles better. Solution: Use the related skills section to hand off deliberately. Keep the imported provenance visible so the next skill inherits the right context instead of starting blind.

Imported Troubleshooting Notes

Imported: Troubleshooting

• Low reply rates → Cause: Generic first lines, no signal-based targeting, or weak CTA. Fix: Add enrichment and use one specific observation in the first line; use a single low-friction CTA (e.g. reply or short call).
• Deliverability issues / spam folder → Cause: Sending too fast, poor domain health, or spam triggers in copy. Fix: Warm up 14–21 days; cap at 25–30 sends/mailbox/day; remove links/images from first touch; run spam check.
• Meetings don’t show up → Cause: CTA is too big (e.g. "book 30 min") or sequence stops too early. Fix: Use lower-friction CTA first (reply, short call); extend to 5–7 touches with 3–5 day spacing.

---

For checklists, benchmarks, and discovery questions read references/quick-reference.md when you need detailed reference.

---

Related Skills

• @accessibility - Use when the work is better handled by that native specialization after this imported skill establishes context.
• @ai-pricing - Use when the work is better handled by that native specialization after this imported skill establishes context.
• @ai-sdr - Use when the work is better handled by that native specialization after this imported skill establishes context.
• @ai-seo - Use when the work is better handled by that native specialization after this imported skill establishes context.

Additional Resources

Use this support matrix and the linked files below as the operator packet for this imported skill. They should reflect real copied source material, not generic scaffolding.

Resource family	What it gives the reviewer	Example path
references	copied reference notes, guides, or background material from upstream	references/benchmarks-deliverability-tactics.md
examples	worked examples or reusable prompts copied from upstream	examples/n/a
scripts	upstream helper scripts that change execution or validation	scripts/n/a
agents	routing or delegation notes that are genuinely part of the imported package	agents/n/a
assets	supporting assets or schemas copied from the source package	assets/n/a

• benchmarks-deliverability-tactics.md
• quick-reference.md
• benchmarks-deliverability-tactics.md
• quick-reference.md

Imported Reference Notes

Imported: The AI Outreach Stack

The modern cold outreach system is a six-stage pipeline. Each stage has specific tools, metrics, and failure modes.

+------------------+     +------------------+     +---------------------+
|  1. SIGNAL       |---->|  2. ENRICHMENT   |---->|  3. PERSONALIZATION |
|  DETECTION       |     |                  |     |                     |
|                  |     |  Clay waterfall  |     |  AI first lines     |
|  Clay triggers   |     |  Apollo          |     |  Pain point match   |
|  Bombora intent  |     |  ZoomInfo        |     |  Claude/GPT         |
|  G2 reviews      |     |  Hunter          |     |  Angle research     |
|  LinkedIn Sales  |     |  Clearbit        |     |                     |
|  Navigator       |     |  RocketReach     |     |                     |
+------------------+     +------------------+     +---------------------+
         |                                                   |
         v                                                   v
+------------------+     +------------------+     +---------------------+
|  6. FOLLOW-UP    |<----|  5. SENDING      |<----|  4. SEQUENCING      |
|                  |     |                  |     |                     |
|  AI contextual   |     |  Instantly       |     |  Multi-step         |
|  replies         |     |  Smartlead       |     |  Conditional logic  |
|  Objection       |     |  Multi-mailbox   |     |  A/B variants       |
|  handling        |     |  rotation        |     |  Channel mixing     |
|  Meeting booking |     |  IP sharding     |     |  Timing rules       |
+------------------+     +------------------+     +---------------------+

Stage 1: Signal Detection

Signals tell you WHO to reach out to and WHEN. Cold email without signals is spam with extra steps.

Signal types ranked by conversion intent:

Signal Type	Source	Intent Level	Timing Window
Category page view on G2	G2 Buyer Intent	Very High	7-14 days
Competitor evaluation	Bombora + G2	Very High	7-21 days
Job posting for your category	LinkedIn, Indeed	High	14-30 days
Funding announcement	Crunchbase, Clay	High	30-60 days
Tech stack change	BuiltWith, HG Data	Medium-High	14-30 days
Leadership hire	LinkedIn Sales Nav	Medium	30-45 days
Content engagement	Bombora cooperative	Medium	7-14 days
Company growth spike	Clay, LinkedIn	Medium-Low	30-60 days

Signal layering strategy: Single signals produce 3-5% reply rates. Layer two or more signals and reply rates jump to 8-15%. Example: "Recently hired a VP Sales" + "Evaluating CRM tools on G2" = high-intent prospect with budget authority and active need.

Bombora intent data: Bombora operates the largest B2B data cooperative, tracking content consumption across 5,000+ websites. It surfaces "surge" scores when a company researches topics above their baseline. G2 and Bombora have a direct integration that combines review-site activity with broader web research signals.

Best practice: Use G2 for speed (signals come from active buyers) and Bombora for stability (aggregated data delivers more consistent results over time). Layer both for full coverage.

Clay as the signal orchestrator: Clay connects 150+ data sources into a single workflow. Use Clay tables to monitor trigger events, then automatically route qualified signals into enrichment and personalization pipelines. Clay's HTTP request action lets you connect any API as a signal source.

Stage 2: Enrichment

Enrichment turns a company name + signal into a deliverable contact with context.

The waterfall enrichment model:

Lead enters Clay table
        |
        v
  [Provider A: Apollo]
  Found email? ----YES----> Verified? --YES--> Done
        |                       |
       NO                      NO
        |                       |
        v                       v
  [Provider B: Hunter]    [Provider C: ZoomInfo]
  Found email? ----YES----> Verified? --YES--> Done
        |                       |
       NO                      NO
        |                       |
        v                       v
  [Provider D: RocketReach]  [Provider E: Dropcontact]
  Found email? ----YES----> Verified? --YES--> Done
        |
       NO
        |
        v
  Skip or manual research

Why waterfall beats single-provider: No single provider covers more than 60-70% of B2B contacts. Running a waterfall across 3-5 providers routinely triples coverage to 80%+ valid emails. Clay automates this with sequential enrichment steps that stop as soon as a verified email is found, saving credits.

Enrichment data to collect (in priority order):

1. Verified work email - Required. Bounce rate must stay under 2%.
2. Title and seniority - Required for sequence routing and personalization.
3. Company size and revenue - Required for ICP filtering.
4. Recent company news - Funding, product launches, expansions. Powers first lines.
5. Tech stack - BuiltWith or HG Data. Critical for displacement plays.
6. LinkedIn profile URL - For multichannel sequences and AI research.
7. Hiring signals - Open roles that indicate pain points or growth.
8. Social posts or articles - Fuel for AI-personalized first lines.

Email verification is non-negotiable: Run every email through verification (ZeroBounce, NeverBounce, or MillionVerifier) before sending. A bounce rate above 2% triggers spam filters at Google and Microsoft. One bad list can burn a domain in a day.

Stage 3: AI Personalization

Generic cold emails get 1-2% reply rates. AI-personalized emails get 8-12%. The difference is the first two lines.

The AI personalization pipeline:

Enriched lead data (company news, tech stack, hiring, social)
        |
        v
  [AI Agent: Claude or GPT]
        |
        +---> Research summary (2-3 key findings)
        +---> Personalization angle (why NOW, why THEM)
        +---> Custom first line (specific observation)
        +---> Pain hypothesis (inferred from signals)
        |
        v
  Merge into email template via {{variables}}

First line frameworks that work:

Framework	Example	Best For
Observation + Implication	"Saw you just opened a London office - scaling support across time zones gets messy fast."	Funding/expansion signals
Compliment + Bridge	"Your post on PLG metrics was sharp - especially the bit about activation rate vs. NPS."	Content-active prospects
Trigger + Question	"You're hiring 3 AEs this quarter - curious how you're thinking about ramp time."	Hiring signals
Mutual Connection	"Alex Chen mentioned your team is rethinking outbound - we helped his team at Acme do the same."	Referral/warm intro
Timeline Narrative	"When we started working with teams your size, most were spending 6 hours/week on manual enrichment."	Timeline hooks (highest reply rate)

Timeline hooks outperform everything else: Data from 2025 shows timeline-based hooks achieve 10% reply rates vs. 4.4% for problem-based hooks - a 2.3x gap. Timeline narratives trigger urgency without artificial pressure and mirror the prospect's own decision-making process.

AI model selection for personalization:

Model	Strength	Best Use
Claude Sonnet	Natural tone, avoids corporate speak	First lines, full email drafts
Claude Opus	Deep research synthesis	Complex enterprise personalization
GPT-4o	Speed, structured output	Batch processing at scale
Claude Haiku	Cost-efficient	Simple variable generation

Claude models produce the most natural-sounding cold emails. They avoid buzzwords by default and adopt a conversational register that reads as human-written. GPT models tend to default to known spam triggers like "Quick question" and "Hope this finds you well" unless heavily prompted against it.

Scaling AI personalization with Clay:

1. Build a Clay table with enriched leads
2. Add an AI enrichment column using Claude
3. Prompt: "Research this company using the data provided. Write a 1-sentence observation about [specific context]. Do not use corporate jargon."
4. Output flows into Instantly/Smartlead as a merge field
5. Cost: roughly $0.01-0.03 per lead for Sonnet-tier models

Stage 4: Sequencing

A sequence is the multi-step campaign structure. It defines how many emails, when they send, and what each email does.

The anatomy of a high-performing sequence:

Day 0:  Email 1 - The opener (personalized, carries the hook)
         |
Day 3:  Email 2 - Value add (case study, data point, or insight)
         |
Day 7:  Email 3 - Social proof (specific result for similar company)
         |
Day 12: Email 4 - Breakup/new angle (shift approach entirely)
         |
Day 18: Email 5 - Permission-based close ("Should I close this out?")

Sequence length and timing rules:

Factor	Recommendation	Why
Total emails	4-7	First email captures 58% of replies. Diminishing returns after 7.
Gap between emails	2-4 business days	3 days is the sweet spot. Less feels pushy, more loses momentum.
Total sequence duration	14-25 days	Beyond 25 days, leads go stale.
SMB sequences	5-8 touches over 30 days	Shorter decision cycles.
Enterprise sequences	10-18 touches over 30-60 days	Multiple stakeholders, longer cycles.

Conditional branching logic: Modern sequences are not linear. Build branches based on:

• Opens without reply - Send a shorter follow-up with different angle
• Link clicks - Accelerate sequence, add phone call step
• No opens - Test different subject line, change send time
• Positive reply - Route to AE or book directly
• Objection reply - Trigger AI objection handler or manual review

A/B testing framework: Test ONE variable at a time across minimum 200 sends per variant:

Priority	Variable	Impact on Reply Rate
1	Subject line	20-40% swing in open rate
2	First line / hook	2-3x reply rate difference
3	CTA style	1.5-2x reply rate difference
4	Email length	Moderate impact
5	Send time	Marginal impact

Stage 5: Sending Infrastructure

Infrastructure is where most outreach systems break. Perfect copy with bad deliverability lands in spam.

Domain and mailbox architecture:

Primary Domain: yourcompany.com
  (NEVER use for cold outreach)

Secondary Domains (for outreach only):
  yourcompany-team.com    --> mailbox1@, mailbox2@
  tryyourcompany.com      --> mailbox1@, mailbox2@
  getyourcompany.com      --> mailbox1@, mailbox2@
  yourcompanyhq.com       --> mailbox1@, mailbox2@

Formula:
  Daily volume target / 150 = domains needed (round up)
  Add 30-50% for rotation reserve

Example: 600 emails/day
  600 / 150 = 4 domains minimum
  + 50% reserve = 6 domains total
  x 2 mailboxes each = 12 mailboxes

Infrastructure sizing guide:

Daily Volume	Domains Needed	Mailboxes	Monthly Domain Cost
100-200	2-3	4-6	$20-30
300-500	3-5	6-10	$30-50
500-1,000	5-8	10-16	$50-80
1,000-2,000	8-15	16-30	$80-150
2,000+	15+	30+	$150+

Per-mailbox sending limits:

Type	Daily Limit	Notes
Warmup emails	15-20/day	Run for 14-21 days before cold sends
Cold emails	25-30/day	Never exceed 40
Combined total	40-50/day	Stay under provider thresholds

Domain warmup protocol:

Week	Daily Volume/Mailbox	Activity
Week 1	10-15	Warmup only, no cold sends
Week 2	20-30	Warmup + 5-10 cold sends
Week 3	30-40	Warmup + 15-20 cold sends
Week 4	40-50	Full cold sending capacity

Authentication setup checklist (do this on Day 1):

• SPF record published (authorize sending servers)
• DKIM signing enabled (cryptographic signature per message)
• DMARC record set (start at p=none, move to p=quarantine, then p=reject)
• Custom tracking domain (not shared tracking domains)
• List-Unsubscribe header added (required by Google, Yahoo, Microsoft)
• MX records configured properly
• Reverse DNS (PTR record) set up