Preamble (runs on skill start)
# Version check (silent if up to date)
python3 telemetry/version_check.py 2>/dev/null || true
# Telemetry opt-in (first run only, then remembers your choice)
python3 telemetry/telemetry_init.py 2>/dev/null || true
Privacy: This skill logs usage locally to ~/.ai-marketing-skills/analytics/. Remote telemetry is opt-in only. No code, file paths, or repo content is ever collected. See telemetry/README.md.
Cold Outbound Optimizer
Startup: Determine Mode
Ask the user:
- Do you have an existing Instantly account with campaigns to audit, or are you starting from scratch?
- Do you have an Instantly API key? (Required for audit mode.)
If API key provided → run scripts/instantly-audit.py to pull campaigns, account inventory, and warmup scores before proceeding.
Phase 1: Discovery & Audit
1A — Infrastructure Check (if API key available)
Run python3 scripts/instantly-audit.py --api-key <KEY> and report:
- Active campaigns (name, status, reply rate, open rate)
- Sending accounts (count, warmup score, daily limit)
- Domain inventory
- Warmup gaps: any account with score <80 or <14 days warmup → flag as NOT ready
1B — Performance Data
- Pull campaign analytics from Instantly
- Ask: "Do you have a spreadsheet with historical outbound data?" If yes, request link.
1C — ICP Definition
If no ICP defined, collect:
- Titles: Who are you targeting? (e.g., VP Marketing, Head of Growth)
- Industries: Which verticals?
- Company size: Employee count or revenue range?
- Revenue floor: Minimum ARR/revenue to qualify?
- Anti-ICP: Who to explicitly exclude?
Use references/icp-template.md as the collection template.
1D — Business Context
Collect:
- What do you sell? (One sentence, no jargon)
- What's the primary offer? (Free trial, audit, demo, consultation)
- Real URLs to reference (pricing page, case studies, relevant content)
- Any proof points? (Client results, stats, social proof)
1E — Expert Panel Config
Default: 10 experts (see references/expert-panel.md).
Ask: "Any industry-specific experts to add, or panelists to swap?" Confirm roster before scoring.
Phase 2: Expert Panel Recursive Scoring
Target: 90/100. Non-negotiable. Iterate until reached.
Round Structure
Each round produces:
- Score table — all 10 panelists, individual score (0-100), one-line rationale
- Aggregate score — average of all 10
- Top weaknesses — ranked list of what's holding the copy back
- Changes made — specific edits addressing each weakness
- Updated copy — full revised sequence after changes
Scoring Criteria (per panelist's lens — see references/expert-panel.md)
- Subject line curiosity / open rate potential
- First sentence pattern interrupt
- Body clarity and brevity
- CTA softness and specificity
- Sequence flow and follow-up logic
- Deliverability risk signals (spam words, link density)
- Personalization believability
Rules
- Scores must be brutally honest. No padding to 90 without earning it.
- If round score < 90: identify top 3 weaknesses, revise copy, run next round.
- If round score ≥ 90: finalize copy and proceed to deliverables.
- Show every round in the final doc — the iteration trail is part of the value.
Phase 3: Deliverables
Strategy Doc
Create a document (Google Doc, Notion, or markdown) with:
- Pre-Analysis / Brutal Truth — what the existing campaigns are doing wrong (or baseline if starting from scratch)
- ICP Summary — confirmed targeting parameters
- Infrastructure Status — account inventory, warmup readiness, capacity math
- Scoring Rounds — full panel vote tables for every round
- Final Email Copy — all steps for all campaigns, Instantly-ready format
- Implementation Plan — step-by-step setup instructions
- Capacity Math — accounts × daily send rate = pipeline projections
- Weekly Metrics Targets — open rate, reply rate, positive reply rate, meetings booked
- STOP List — what to kill immediately
- START List — what to launch first
Format Rules for Final Copy
Follow all rules in references/instantly-rules.md and references/copy-rules.md.
Human Review Gate
Do NOT push anything to Instantly automatically. The doc is for human review. Get explicit approval before any API writes.
Iteration
After review, collect feedback and re-run scoring on revised copy if needed.
Capacity Math Formula
Accounts ready (score ≥80, ≥14 days warmup) × 30 emails/day = conservative daily volume
Accounts ready × 50 emails/day = aggressive daily volume
Daily volume × 22 working days = monthly send capacity
Monthly sends × expected reply rate = expected replies
Expected replies × qualification rate = pipeline opportunities
Weekly Metrics Targets (Baselines)
| Metric |
Good |
Great |
| Open rate |
40%+ |
60%+ |
| Reply rate |
3%+ |
7%+ |
| Positive reply rate |
1%+ |
3%+ |
| Meeting rate |
0.5%+ |
1.5%+ |
Adjust targets based on niche and offer. Cold traffic to a free audit converts differently than a paid trial.
Add-On Recommendations (mention but don't build)
- LinkedIn automation: HeyReach or similar for multi-channel sequences. Separate workflow.
- Lead enrichment: Clay or Apollo for personalization data before upload.
- Lead pipeline: Use
scripts/lead-pipeline.py for Apollo → LeadMagic → Instantly automation.
Reference Files
| File |
Purpose |
references/instantly-rules.md |
Variable syntax, sequence structure, deliverability rules |
references/expert-panel.md |
Default 10-expert roster with scoring lenses |
references/copy-rules.md |
Email copy rules (first sentence, CTA, stats framing) |
references/icp-template.md |
ICP data collection template |
scripts/instantly-audit.py |
Pulls campaigns, accounts, warmup scores via Instantly v2 API |
scripts/lead-pipeline.py |
End-to-end lead sourcing pipeline |
scripts/competitive-monitor.py |
Competitor tracking and intelligence |
scripts/cross-signal-detector.py |
Multi-source signal detection |
scripts/cold-outbound-sender.py |
Send approved outbound emails |
1---
2name: outbound-engine
3description: Design, analyze, and optimize cold outbound email campaigns for Instantly. Handles end-to-end ICP definition, expert panel scoring (recursive to 90+), sequence copywriting, infrastructure audit, capacity planning, and implementation docs. Use when asked to build cold outbound sequences, optimize cold email, analyze outbound campaigns, build sales sequences, build Instantly sequences, create cold outbound strategies, or design email campaigns. Supports both "start from scratch" and "optimize existing" modes.
4description_zh: "自动化外拓邮件引擎,从 ICP 定义到邮件入站全流程"
5description_en: "Automated cold outbound engine from ICP definition to inbox delivery"
6version: 1.0.0
7homepage: https://github.com/ericosiu/ai-marketing-skills
8metadata:
9 source: https://skillmd.com/skills/bighardperson/outbound-engine
10---
11
12
13## Preamble (runs on skill start)
14
15```bash
16# Version check (silent if up to date)
17python3 telemetry/version_check.py 2>/dev/null || true
18
19# Telemetry opt-in (first run only, then remembers your choice)
20python3 telemetry/telemetry_init.py 2>/dev/null || true
21```
22
23> **Privacy:** This skill logs usage locally to `~/.ai-marketing-skills/analytics/`. Remote telemetry is opt-in only. No code, file paths, or repo content is ever collected. See `telemetry/README.md`.
24
25---
26
27# Cold Outbound Optimizer
28
29---
30
31## Startup: Determine Mode
32
33Ask the user:
341. Do you have an **existing Instantly account** with campaigns to audit, or are you **starting from scratch**?
352. Do you have an **Instantly API key**? (Required for audit mode.)
36
37If API key provided → run `scripts/instantly-audit.py` to pull campaigns, account inventory, and warmup scores before proceeding.
38
39---
40
41## Phase 1: Discovery & Audit
42
43### 1A — Infrastructure Check (if API key available)
44Run `python3 scripts/instantly-audit.py --api-key <KEY>` and report:
45- Active campaigns (name, status, reply rate, open rate)
46- Sending accounts (count, warmup score, daily limit)
47- Domain inventory
48- Warmup gaps: any account with score <80 or <14 days warmup → flag as NOT ready
49
50### 1B — Performance Data
51- Pull campaign analytics from Instantly
52- Ask: "Do you have a spreadsheet with historical outbound data?" If yes, request link.
53
54### 1C — ICP Definition
55If no ICP defined, collect:
56- **Titles:** Who are you targeting? (e.g., VP Marketing, Head of Growth)
57- **Industries:** Which verticals?
58- **Company size:** Employee count or revenue range?
59- **Revenue floor:** Minimum ARR/revenue to qualify?
60- **Anti-ICP:** Who to explicitly exclude?
61
62Use `references/icp-template.md` as the collection template.
63
64### 1D — Business Context
65Collect:
66- What do you sell? (One sentence, no jargon)
67- What's the primary offer? (Free trial, audit, demo, consultation)
68- Real URLs to reference (pricing page, case studies, relevant content)
69- Any proof points? (Client results, stats, social proof)
70
71### 1E — Expert Panel Config
72Default: 10 experts (see `references/expert-panel.md`).
73Ask: "Any industry-specific experts to add, or panelists to swap?" Confirm roster before scoring.
74
75---
76
77## Phase 2: Expert Panel Recursive Scoring
78
79**Target: 90/100. Non-negotiable. Iterate until reached.**
80
81### Round Structure
82Each round produces:
831. **Score table** — all 10 panelists, individual score (0-100), one-line rationale
842. **Aggregate score** — average of all 10
853. **Top weaknesses** — ranked list of what's holding the copy back
864. **Changes made** — specific edits addressing each weakness
875. **Updated copy** — full revised sequence after changes
88
89### Scoring Criteria (per panelist's lens — see `references/expert-panel.md`)
90- Subject line curiosity / open rate potential
91- First sentence pattern interrupt
92- Body clarity and brevity
93- CTA softness and specificity
94- Sequence flow and follow-up logic
95- Deliverability risk signals (spam words, link density)
96- Personalization believability
97
98### Rules
99- Scores must be brutally honest. No padding to 90 without earning it.
100- If round score < 90: identify top 3 weaknesses, revise copy, run next round.
101- If round score ≥ 90: finalize copy and proceed to deliverables.
102- Show every round in the final doc — the iteration trail is part of the value.
103
104---
105
106## Phase 3: Deliverables
107
108### Strategy Doc
109Create a document (Google Doc, Notion, or markdown) with:
110
1111. **Pre-Analysis / Brutal Truth** — what the existing campaigns are doing wrong (or baseline if starting from scratch)
1122. **ICP Summary** — confirmed targeting parameters
1133. **Infrastructure Status** — account inventory, warmup readiness, capacity math
1144. **Scoring Rounds** — full panel vote tables for every round
1155. **Final Email Copy** — all steps for all campaigns, Instantly-ready format
1166. **Implementation Plan** — step-by-step setup instructions
1177. **Capacity Math** — accounts × daily send rate = pipeline projections
1188. **Weekly Metrics Targets** — open rate, reply rate, positive reply rate, meetings booked
1199. **STOP List** — what to kill immediately
12010. **START List** — what to launch first
121
122### Format Rules for Final Copy
123Follow all rules in `references/instantly-rules.md` and `references/copy-rules.md`.
124
125### Human Review Gate
126**Do NOT push anything to Instantly automatically.** The doc is for human review. Get explicit approval before any API writes.
127
128### Iteration
129After review, collect feedback and re-run scoring on revised copy if needed.
130
131---
132
133## Capacity Math Formula
134
135```
136Accounts ready (score ≥80, ≥14 days warmup) × 30 emails/day = conservative daily volume
137Accounts ready × 50 emails/day = aggressive daily volume
138Daily volume × 22 working days = monthly send capacity
139Monthly sends × expected reply rate = expected replies
140Expected replies × qualification rate = pipeline opportunities
141```
142
143---
144
145## Weekly Metrics Targets (Baselines)
146
147| Metric | Good | Great |
148|--------|------|-------|
149| Open rate | 40%+ | 60%+ |
150| Reply rate | 3%+ | 7%+ |
151| Positive reply rate | 1%+ | 3%+ |
152| Meeting rate | 0.5%+ | 1.5%+ |
153
154Adjust targets based on niche and offer. Cold traffic to a free audit converts differently than a paid trial.
155
156---
157
158## Add-On Recommendations (mention but don't build)
159
160- **LinkedIn automation:** HeyReach or similar for multi-channel sequences. Separate workflow.
161- **Lead enrichment:** Clay or Apollo for personalization data before upload.
162- **Lead pipeline:** Use `scripts/lead-pipeline.py` for Apollo → LeadMagic → Instantly automation.
163
164---
165
166## Reference Files
167
168| File | Purpose |
169|------|---------|
170| `references/instantly-rules.md` | Variable syntax, sequence structure, deliverability rules |
171| `references/expert-panel.md` | Default 10-expert roster with scoring lenses |
172| `references/copy-rules.md` | Email copy rules (first sentence, CTA, stats framing) |
173| `references/icp-template.md` | ICP data collection template |
174| `scripts/instantly-audit.py` | Pulls campaigns, accounts, warmup scores via Instantly v2 API |
175| `scripts/lead-pipeline.py` | End-to-end lead sourcing pipeline |
176| `scripts/competitive-monitor.py` | Competitor tracking and intelligence |
177| `scripts/cross-signal-detector.py` | Multi-source signal detection |
178| `scripts/cold-outbound-sender.py` | Send approved outbound emails |