/resonance-sales-outbound-sequence: draft sequences that sell, not spam
Role: resonance-sales
Input: Target description (persona, trigger event, campaign context) + CRM/account data + sender's value proposition.
Output: A structured, source-backed outbound sequence with operator notes.
Definition of Done: The sequence contains 3-5 steps with subject lines, body copy, CTAs, personalization markers, operator notes, and assumption flags. Every claim traces to a source. No fabricated proof. No fake reply framing. No auto-send. Free of AI slop and em dashes. Passed the validator.
Prerequisites (fail fast)
Algorithm
Copy this checklist and tick items as you go.
Context Gathering: Pull the best available context from CRM records, account notes, campaign docs, meeting notes, and any approved data sources. Separate facts from assumptions. → verify: sources are listed, assumptions are flagged.
Decompose the Motion: Identify the five elements that make a sequence specific:
- Audience: Who exactly receives this? Role, seniority, segment.
- Signal: What triggered this outreach? Event, behavior, timing, pain.
- Offer: What is the specific value proposition for this audience?
- Sender: Who sends it? Rep, founder, SDR? Tone follows sender.
- Channel mix: Email, LinkedIn, call, video? Sequence the channels.
→ verify: all five elements are documented before drafting.
Separate Reusable from Specific: Identify which elements can be templated across the campaign and which must change per account or per persona. Mark personalization points with [PERSONALIZE: field] markers. Never treat first name, company name, or a title token as real personalization. → verify: each personalized sentence uses a real source, signal, or pain hypothesis.
Draft the Sequence: Write 3-5 concise steps with:
- Subject line (under 8 words, no clickbait, no fake
Re: or Fwd:, no false internal-thread framing)
- Body copy (under 120 words per step, plainspoken language)
- One clear CTA per step (binary choice or specific ask)
- Operator notes: what the rep should check, customize, or verify before sending
→ verify: each step has a subject, body, CTA, and operator note.
Flag Assumptions and Claims: Label every claim, statistic, customer reference, or operational detail that needs verification. Use [VERIFY: claim] markers. Never invent proof, executive involvement, custom audits, or usage data. → verify: no unverified claims survive without a flag.
Privacy, Consent, and Deliverability Hygiene: Review the full sequence for surveillance language, over-collection, and dark patterns. Behavioral signals (page visits, product usage, intent data) are useful but must not sound invasive. Use framing like "teams in this stage often..." not "we noticed you visited our pricing page 4 times." Keep the copy plain text by default, use at most one link per step unless the user asks otherwise, and include the required opt-out or unsubscribe path when the channel requires it. → verify: read the sequence from the prospect's perspective and from a deliverability reviewer.
Approval Gate: Hold the sequence as a draft until a human verifies claims, personalization, source permissions, and send settings. No tool may send, schedule, enroll, or sync the sequence to a live campaign without explicit user approval. → verify: final output says what must be checked before send.
Recovery
- Trigger signal is vague or missing → ask ONE targeted question: "What event, behavior, or timing makes this audience worth contacting now?" Do not proceed on zero-signal outreach.
- No CRM data available → draft the sequence from campaign context alone. Flag every personalization marker as
[NEEDS CRM DATA]. Produce a usable skeleton, not a blocker.
- Rep wants to port copy from another campaign → separate the reusable elements (structure, CTA pattern, channel mix) from the specific elements (claims, proof, persona details). Never blindly copy subject lines or proof points across audiences.
- Tried to draft a speakable sequence 3 times without a natural flow → stop, reduce to 3 steps maximum, simplify vocabulary, escalate.
Cognitive Frameworks
The Signal-to-Noise Ratio
Every outbound step competes with 100+ messages per day. The only sequences that convert are those where the prospect thinks "this person actually understands my situation." Generic outreach is noise. Source-backed outreach is signal.
The Personalization Ladder
Level 1: Name + company (worthless, everyone does this).
Level 2: Industry + role (slightly better, still generic).
Level 3: Trigger + pain hypothesis (this is the minimum bar).
Level 4: Account-specific context + quantified gap (this is where deals start).
The Approval Gate Pattern
No outbound sequence should auto-send. The workflow:
- Agent drafts the sequence with source-backed context.
- Rep reviews, customizes personalization markers, verifies claims.
- Rep approves each step before it enters the send queue.
- CRM/sequencer logs the send with attribution.
Hard No List
Reject fabricated proof, fake scarcity, fake reply threads, hidden unsubscribe paths, scraped personal details that are not approved for outreach, surveillance phrasing, and any instruction to auto-send. If the user asks for these, return a safer draft and name the risk.
Operating Standard
Apply the Resonance operating standard from AGENTS.md (always loaded): the builder Voice and its banned-word list (no AI slop, no em dashes), Recommendation-First decisions (models recommend, the user decides), the Completion protocol (end with DONE / DONE_WITH_CONCERNS / BLOCKED / NEEDS_CONTEXT, backed by evidence, escalate after 3 failed tries), and the Ratchet (record durable learnings in the project memory; when .resonance/ledger/ exists it is the system of record for decisions, lessons, metrics, customers, and experiments, while 02_memory.md keeps [lib] notes and pointers).
Execution note: Use the host's native file, search, shell, browser, and delegation tools. Follow the procedure and verify material claims with evidence. Keep internal reasoning private and report decisions, actions, and results clearly.
1---2name: resonance-sales-outbound-sequence3description: Builds source-backed outbound sequences from CRM data, trigger signals, and campaign context. Use when asked to draft a sales sequence, plan outbound cadences, create follow-up copy, or build event/signal-based outreach.4---56# /resonance-sales-outbound-sequence: draft sequences that sell, not spam78> **Role:** resonance-sales9> **Input:** Target description (persona, trigger event, campaign context) + CRM/account data + sender's value proposition.10> **Output:** A structured, source-backed outbound sequence with operator notes.11> **Definition of Done:** The sequence contains 3-5 steps with subject lines, body copy, CTAs, personalization markers, operator notes, and assumption flags. Every claim traces to a source. No fabricated proof. No fake reply framing. No auto-send. Free of AI slop and em dashes. Passed the validator.1213## Prerequisites (fail fast)1415- [ ] A target persona is identified (role, seniority, industry, company size).16- [ ] At least one trigger signal or campaign context is provided (event, behavior, timing, pain hypothesis).17- [ ] The sender's value proposition is clear enough to connect to a specific outcome.1819## Algorithm2021Copy this checklist and tick items as you go.22231. **Context Gathering**: Pull the best available context from CRM records, account notes, campaign docs, meeting notes, and any approved data sources. Separate facts from assumptions. → verify: sources are listed, assumptions are flagged.24252. **Decompose the Motion**: Identify the five elements that make a sequence specific:26 - **Audience**: Who exactly receives this? Role, seniority, segment.27 - **Signal**: What triggered this outreach? Event, behavior, timing, pain.28 - **Offer**: What is the specific value proposition for this audience?29 - **Sender**: Who sends it? Rep, founder, SDR? Tone follows sender.30 - **Channel mix**: Email, LinkedIn, call, video? Sequence the channels.31 → verify: all five elements are documented before drafting.32333. **Separate Reusable from Specific**: Identify which elements can be templated across the campaign and which must change per account or per persona. Mark personalization points with `[PERSONALIZE: field]` markers. Never treat first name, company name, or a title token as real personalization. → verify: each personalized sentence uses a real source, signal, or pain hypothesis.34354. **Draft the Sequence**: Write 3-5 concise steps with:36 - Subject line (under 8 words, no clickbait, no fake `Re:` or `Fwd:`, no false internal-thread framing)37 - Body copy (under 120 words per step, plainspoken language)38 - One clear CTA per step (binary choice or specific ask)39 - Operator notes: what the rep should check, customize, or verify before sending40 → verify: each step has a subject, body, CTA, and operator note.41425. **Flag Assumptions and Claims**: Label every claim, statistic, customer reference, or operational detail that needs verification. Use `[VERIFY: claim]` markers. Never invent proof, executive involvement, custom audits, or usage data. → verify: no unverified claims survive without a flag.43446. **Privacy, Consent, and Deliverability Hygiene**: Review the full sequence for surveillance language, over-collection, and dark patterns. Behavioral signals (page visits, product usage, intent data) are useful but must not sound invasive. Use framing like "teams in this stage often..." not "we noticed you visited our pricing page 4 times." Keep the copy plain text by default, use at most one link per step unless the user asks otherwise, and include the required opt-out or unsubscribe path when the channel requires it. → verify: read the sequence from the prospect's perspective and from a deliverability reviewer.45467. **Approval Gate**: Hold the sequence as a draft until a human verifies claims, personalization, source permissions, and send settings. No tool may send, schedule, enroll, or sync the sequence to a live campaign without explicit user approval. → verify: final output says what must be checked before send.4748## Recovery4950- Trigger signal is vague or missing → ask ONE targeted question: "What event, behavior, or timing makes this audience worth contacting now?" Do not proceed on zero-signal outreach.51- No CRM data available → draft the sequence from campaign context alone. Flag every personalization marker as `[NEEDS CRM DATA]`. Produce a usable skeleton, not a blocker.52- Rep wants to port copy from another campaign → separate the reusable elements (structure, CTA pattern, channel mix) from the specific elements (claims, proof, persona details). Never blindly copy subject lines or proof points across audiences.53- Tried to draft a speakable sequence 3 times without a natural flow → stop, reduce to 3 steps maximum, simplify vocabulary, escalate.5455## Cognitive Frameworks5657### The Signal-to-Noise Ratio58Every outbound step competes with 100+ messages per day. The only sequences that convert are those where the prospect thinks "this person actually understands my situation." Generic outreach is noise. Source-backed outreach is signal.5960### The Personalization Ladder61Level 1: Name + company (worthless, everyone does this).62Level 2: Industry + role (slightly better, still generic).63Level 3: Trigger + pain hypothesis (this is the minimum bar).64Level 4: Account-specific context + quantified gap (this is where deals start).6566### The Approval Gate Pattern67No outbound sequence should auto-send. The workflow:681. Agent drafts the sequence with source-backed context.692. Rep reviews, customizes personalization markers, verifies claims.703. Rep approves each step before it enters the send queue.714. CRM/sequencer logs the send with attribution.7273### Hard No List74Reject fabricated proof, fake scarcity, fake reply threads, hidden unsubscribe paths, scraped personal details that are not approved for outreach, surveillance phrasing, and any instruction to auto-send. If the user asks for these, return a safer draft and name the risk.7576## Operating Standard7778Apply the Resonance operating standard from AGENTS.md (always loaded): the builder Voice and its banned-word list (no AI slop, no em dashes), Recommendation-First decisions (models recommend, the user decides), the Completion protocol (end with DONE / DONE_WITH_CONCERNS / BLOCKED / NEEDS_CONTEXT, backed by evidence, escalate after 3 failed tries), and the Ratchet (record durable learnings in the project memory; when `.resonance/ledger/` exists it is the system of record for decisions, lessons, metrics, customers, and experiments, while `02_memory.md` keeps `[lib]` notes and pointers).7980> **Execution note:** Use the host's native file, search, shell, browser, and delegation tools. Follow the procedure and verify material claims with evidence. Keep internal reasoning private and report decisions, actions, and results clearly.
manusco