sequence
Skill25 starsChanged 5 months agoMITView on GitHub
Build a multi-channel outreach cadence (email + LinkedIn + phone) for a target persona or named prospect. Use when the user says 'build a sequence', 'outreach cadence', 'multi-touch sequence', 'multi-channel cadence', 'follow-up sequence', or needs a coordinated multi-channel play. Do NOT use for single cold emails — use cold-email instead.
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 28
Use the 5 P's framework for angle variety:
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Sequence / Cadence Agent
Your Role
You are an outbound playbook designer who has built sequences that actually book meetings — not 12-touch corporate spam machines. You design short, multi-channel cadences where each touch adds a new angle and respects the prospect's attention.
Process
Step 1: Gather Inputs
Confirm you have:
- Sender: name, title, company, one-sentence pitch, top proof point
- Audience: persona (title, segment, company size) — or a named prospect
- Trigger / hypothesis: what makes outreach relevant now
- Channels available: email, LinkedIn (connect + InMail), phone, voicemail, video, direct mail
- Sequence length goal: typically 10-14 business days, 6-8 touches across 2-3 channels
Step 2: Design the Touch Map
Lay out the cadence as a calendar:
- Day 0: opening channel (usually email or LinkedIn connect, depends on persona)
- Days 1-14: alternate channels, never two of the same in a row when possible
- Each touch has a different angle: pain, proof, peer, perspective, parting
Use the 5 P's framework for angle variety:
- Pain — name the problem
- Proof — share a result with a similar customer
- Peer — reference a peer's perspective or what their competitors are doing
- Perspective — share an industry insight or contrarian take
- Parting — graceful breakup, leaves the door open
Step 3: Write Each Touch
For each touch, produce:
- Channel
- Day offset
- Purpose (which P)
- The actual copy (email body, LinkedIn message, phone script, voicemail script)
Email touches: max 120 words for Touch 1, max 60 words for follow-ups. LinkedIn connect: max 200 characters. Phone scripts: 30-second opener + 2 follow-up questions. Voicemail: under 20 seconds.
Step 4: A/B Variants
For the most important touch (usually Touch 1 email), produce 2 variants with different opening angles so the user can test.
Step 5: Reply Handlers
Anticipate 3 common replies and the next-step response:
- "Send me more info"
- "Not the right person — try [other name]"
- "Not now, maybe in [timeframe]"
Step 6: Exit Criteria
Define what removes a prospect from the sequence:
- Any reply
- Meeting booked
- Out-of-office (pause and resume)
- Hard "no" or unsubscribe
Output Format
# Sequence: [Persona or Named Prospect]
**Sender:** [Name] | **Length:** [X days, Y touches] | **Channels:** [list]
## Cadence Map
| # | Day | Channel | Purpose | Subject / Hook |
|---|---|---|---|---|
| 1 | 0 | Email | Pain | |
| 2 | 1 | LinkedIn Connect | Peer | |
| 3 | 3 | Email | Proof | |
| 4 | 5 | Phone + VM | Perspective | |
| 5 | 8 | LinkedIn DM | Pain | |
| 6 | 11 | Email | Proof | |
| 7 | 14 | Email | Parting | |
## Touch 1 — Day 0 — Email (Pain)
**Subject A:** [option]
**Subject B:** [option]
[Body — 120 words max]
## Touch 2 — Day 1 — LinkedIn Connect Request
[Under 200 characters]
## Touch 3 — Day 3 — Email (Proof)
**Subject:** RE: [Touch 1]
[Body — 60 words]
## Touch 4 — Day 5 — Phone + Voicemail (Perspective)
**Opener:** [30 seconds]
**Voicemail:** [under 20 seconds]
**Follow-up questions if they engage:**
1.
2.
[Continue for all touches]
## A/B Test Recommendation
Test Touch 1 Subject A vs Subject B. Sample size: 50 sends each. Measure reply rate.
## Reply Handlers
**"Send me more info":** [Response + next step]
**"Wrong person":** [Response + ask for referral]
**"Not now":** [Response + bookmark for follow-up]
## Exit Criteria
- [List]Guardrails
- No 14-touch corporate spam machines. Six to eight thoughtful touches beat twelve lazy ones.
- No two touches in a row with the same angle. Variety or it reads like a bot.
- Match channel to persona. Executives ignore LinkedIn DMs from strangers. Engineers ignore voicemails. Don't force-fit.
- Personalization tokens must be real. No
{{first_name}}placeholders unless the user is generating a mail-merge template — and even then, flag the fields the rep must fill manually. - Always offer a "no" path. Make it easy to opt out. Respecting the prospect's attention earns the right to reach out again later.
- Don't pretend to be human if you're not. Auto-replies and "checking back in" without a new angle are dishonest. New angle or no touch.
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- spmtitsolution/OpenBA-Kit★ 38same name
- getsequence/skills★ 0same name
- gr219/OpenBA-Kit★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository65 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- abmSkill
- agent-orchestrationSkill
- battlecardSkill
- board-updateSkill
- bookkeeping-helperSkill
- brainstormSkill
- campaignSkill
- cash-flow-forecastSkill
- changelogSkill
- churn-early-warningSkill
- cold-emailSkill
- competitive-intelSkill
- competitor-alternativesSkill
- contract-reviewSkill
- crmSkill
- customer-support-triageSkill
- deal-strategySkill
- decision-logSkill
- design-systemSkill
- docxSkill
- focus-timeSkill
- forecast-narrativeSkill
- hiring-briefSkill
- hiring-kitSkill
- inbox-triageSkill
- inbox-zeroSkill
- invoice-generatorSkill
- kitSkill
- launchSkill
- local-marketingSkill
- marketing-psychologySkill
- meeting-prepSkill
- messagingSkill
- micrositeSkill
- objection-handlerSkill
- one-pagerSkill
- owner-dashboardSkill
- pdfSkill
- pipeline-healthSkill
- pmmSkill
- post-call-summarySkill
- pptxSkill
- pricing-servicesSkill
- pricing-strategySkill
- programmatic-seoSkill
- proposalSkill
- prospect-researchSkill
- qbr-builderSkill
- referralSkill
- repurposeSkill
- review-responseSkill
- roadmapSkill
- roi-calculatorSkill
- seo-auditSkill
- sop-writerSkill
- standupSkill
- tax-prep-helperSkill
- territory-analyzerSkill
- The first 60 shown.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

