linkedin-prospecting
Skill1 starsChanged 2 months agoMITView on GitHub
Use LinkedIn for sourcing prospects and running multichannel outreach alongside email, within the platform's limits. Use whenever the user mentions LinkedIn prospecting, Sales Navigator, connection requests, InMail, LinkedIn automation, social selling, multichannel outreach, or combining LinkedIn with cold email. Also use when email-only outreach has plateaued.
In this repository30 other files it gives its agents.
- agency-multi-client-outboundSkill
- bounce-recoverySkill
- catch-all-risky-emailsSkill
- cold-email-copywritingSkill
- crm-data-hygieneSkill
- ecommerce-store-prospectingSkill
- email-deliverabilitySkill
- email-sequence-designSkill
- email-verificationSkill
- email-warmupSkill
- funding-news-signalsSkill
- google-maps-lead-scrapingSkill
- icp-definitionSkill
- inbox-rotation-scalingSkill
- job-posting-signalsSkill
- lead-enrichmentSkill
- lead-scoring-qualificationSkill
- lead-sourcingSkill
- meeting-booking-ctaSkill
- outbound-campaign-opsSkill
- outbound-complianceSkill
- outbound-metrics-benchmarksSkill
- personalization-at-scaleSkill
- reply-handlingSkill
- spam-filter-triggersSkill
- spf-dkim-dmarc-setupSkill
- subject-line-optimizationSkill
- suppression-list-managementSkill
- technographic-signalsSkill
- web-scraping-lead-dataSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
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 14
Current titles: Every other source decays.
- Distinctive: line 44
Automating LinkedIn violates the user agreement: Not criminal, but the platform enforces it and the penalty is the account.
- Distinctive: line 78
If you write one, keep it under 200 characters and make it a reason, not a pitch:
- Distinctive: line 90
Ask about their situation rather than describing yours.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
LinkedIn Prospecting
LinkedIn is the best source of current job titles and the worst channel to automate aggressively. Both facts follow from the same thing: it's a walled platform that actively defends itself, and account restrictions are permanent enough to matter.
Used as research plus a light touch alongside email, it lifts reply rates meaningfully. Used as an automation channel at volume, it gets accounts restricted and the sender learns nothing.
What it's genuinely best at
Current titles. Every other source decays; LinkedIn is updated by the person themselves. For confirming you have the right person in the right role right now, nothing else comes close.
Finding the right person. Filtering a company's employees by function and seniority is faster and more accurate than guessing from a website.
Job change detection. Someone leaving is a dead lead at the old company and a warm one at the new. Someone new in a buying role is one of the strongest triggers there is. See funding-news-signals.
Research material. Recent posts, comments, shared articles and profile changes are dated, public and specific — good material for a first line, if used carefully.
Warm paths. Shared connections, groups, alumni. A genuine mutual connection changes everything about a first message.
The limits, and they're real
Approximate daily ceilings for a normal account. They vary by account age, standing and whether you're on a paid tier, and the platform doesn't publish them.
| Action | Rough safe ceiling |
| Connection requests | 15-20/day, 100/week |
| Profile views | 100-150/day |
| Messages to connections | 30-50/day |
| InMail | Per your plan's allowance |
| Follows | 30-50/day |
Exceeding them produces warnings, then temporary restrictions, then permanent ones. A restricted account loses the network it took years to build, and appeals mostly fail.
Connection acceptance rate matters more than volume. A low acceptance rate signals spraying, and the platform notices. Twenty targeted requests at 40% acceptance is better in every way than a hundred at 8%.
On automation tools
Be straight with users about this rather than either promoting or refusing.
- Automating LinkedIn violates the user agreement. Not criminal, but the platform enforces it and the penalty is the account.
- Browser-extension tools operating in your session are harder to detect than cloud tools logging in from a datacentre IP, which are detected routinely.
- Detection has improved and continues to. Tools that were safe last year aren't now.
- The cost is asymmetric. Modest efficiency gain against losing an account you can't rebuild.
The defensible middle: use LinkedIn manually for research and a small number of genuine touches, and run volume through email where you own the infrastructure. If someone insists on automation, keep it well under the manual limits, use a browser-based tool, warm the account slowly, and accept the risk knowingly.
The multichannel sequence
The pattern that works, and it's mostly email with LinkedIn as reinforcement.
Day 0 Email 1 The pitch, personalised
Day 1 LinkedIn: view profile Passive. They often see it
Day 3 Email 2 New angle
Day 5 LinkedIn: connection request No note, or one short line
Day 7 Email 3 Value, no ask
Day 10 LinkedIn: message Only if they connected
Day 14 Email 4 Close-outWhy this order: email does the work, LinkedIn provides recognition. Someone who's seen your name three times in different places responds differently from someone seeing a cold email once.
Never send the same message on both channels. It reads as automated because it is.
Stop everything on a reply from either channel. Continuing on LinkedIn after they replied to your email is the most common multichannel failure and it undoes the goodwill immediately.
Connection requests
With a note or without? Both work and the data is genuinely mixed.
- No note often has a higher acceptance rate. Low friction, nothing to object to.
- A short note produces lower acceptance but better-qualified acceptance.
If you write one, keep it under 200 characters and make it a reason, not a pitch:
Saw your post on outbound routing — we hit exactly that problem last year. Following your work.
What doesn't work: pitching in the connection request. It's the fastest route to being ignored and reported, and it makes the acceptance useless anyway because they now expect a sales message.
Messaging connections
Once connected, the message is a cold email with a shorter format. Same rules: specific opener, one small ask, no wall of text, no pitch deck link.
Thanks for connecting. Quick one — you mentioned in your post that the SDR ramp was the bottleneck. Is that still the case, or did you solve it?
Ask about their situation rather than describing yours. A question about them gets answered; a paragraph about you doesn't.
Research without automation
Manual research is slow but it's the tier-1 play for your best leads. What's worth reading:
- Recent posts and comments. Dated, public, specific. The best first-line material on the platform.
- The about section, where people describe what they own in their own words.
- Tenure. Someone three months into a role is a very different prospect from someone at eight years.
- Previous companies. A path from a company that used your product is a warm angle.
- What they engage with, which tells you what they care about right now.
Referencing a post is strong. Referencing something private, or something that requires having studied their profile closely, tips into uncomfortable. The test: would they be pleased or unsettled that you knew this? Public post, pleased. Detailed career history recited back at them, unsettled.
Getting emails from LinkedIn identification
The common workflow: identify the right person on LinkedIn, then find their work email elsewhere and run the sequence over email where you control the infrastructure.
That's a sound approach, with one requirement — the address still has to be verified before sending. Pattern-inferred addresses from a name and a domain bounce at rates that damage domains. See email-verification.
Don't do these
- Don't automate at volume. The downside is permanent.
- Don't pitch in a connection request.
- Don't send identical content on both channels.
- Don't message everyone who accepts. Some connections are worth having and not selling to.
- Don't scrape profile data in ways the user agreement prohibits, and be honest with clients about that exposure.
- Don't treat a profile view as a touch worth counting.
- Don't continue on LinkedIn after an email reply.
- Don't build the programme on LinkedIn. You don't own the channel and it can be taken away.
Related skills
lead-sourcing, email-sequence-design, email-verification, personalization-at-scale, funding-news-signals, outbound-compliance
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.

