bereach-ai/linkedin-prospecting-skills / skills/linkedin-messages
linkedin-messages
VerifiedWrites LinkedIn invitation notes, first messages and follow-ups: one template per persona, then per lead a real fact framing its offer. Use when messages sound generic or repeat the title.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 7.8 KB |
| references/artifacts.md | markdown | 8.7 KB |
| references/first-dm.md | markdown | 2.4 KB |
| references/follow-ups.md | markdown | 3.0 KB |
| references/frameworks.md | markdown | 7.4 KB |
| references/invitation-note.md | markdown | 1.1 KB |
| references/sequence-5.md | markdown | 389 B |
references/frameworks.md
View on GitHub# Reference frameworks for LinkedIn messages
**How to use this file.** No framework has been proven better than another: controlled tests find message
design choices make little difference (O'Keefe and Hoeken 2021), and nobody has compared two frameworks on
LinkedIn messages. What is measured is the parts: a reason about them, their problem rather than a pitch,
one easy ask. So SKILL.md gives each message type ONE framework. A test arm changes one part (the
linkedin-outreach-method skill's `references/measuring.md`, Choosing the next test).
**The reader's trade.** Name what is practice in the reader's trade only from these skills or a sourced web
page Claude has read, and a gap is named as a gap ("aucune étude n'a mesuré les messages d'éditeurs de logiciels de calibration à des
responsables réseau"), never as "general rules apply".
**Cite honestly.** Each verdict rests on a principle, a peer-reviewed study or a regulator text from this
file or a source Claude has read, named ("réciprocité", "spécificité", Sahni et al.). A vendor figure (the 85M-email report included)
is labelled "chiffre d'éditeur", told once with its grade in plain words, and only as colour: never "le
mieux mesuré", never "la mieux étayée". Never "tous les experts" without naming them. The CNIL only with
its own conditions (who writes and why this reader, an easy refusal) and its caveat that its page does not
name social network messaging: "la CNIL demande, pour la prospection B2B, de dire qui écrit et pourquoi ce
lecteur ; sa page ne cite pas les messageries des réseaux sociaux, on l'applique par prudence". It is met by
the message's own parts (who writes, the reason, a yes/no ask): never a legal, opt-out or data-source line
in a LinkedIn message. Replies
count only when engaged (a question, a yes, a conversation), and a reply count is never "works because".
Evidence words: **Measured** (study or large dataset with a stated method), **Survey** (self-reported),
**Practice** (named practitioners, nobody measured it), **Vendor claim** (a seller's own figures),
**Untraced** (repeated, no primary source found).
## The eight cards
All examples: one fictional seller, Paul, field calibration software for water networks; every figure is
illustrative.
### Premise first (Becc Holland, Flip the Script)
- What it says: the premise (why you, why now) comes first; the value is mapped onto it; one short ask.
Premises by strength: what they wrote, what they engaged with, how they describe themselves, company
news. With none, a line true for the whole persona (the moment line); never a fake premise.
- Use when / not when: every first message and every invitation note; never with an invented premise.
- Example: "I read your post on winter drift, and a drifted station usually means a second trip."
- Evidence: Practice. Activity-based reasons replied most in email data (vendor).
### Levels of awareness (Eugene Schwartz, *Breakthrough Advertising*, 1966)
- What it says: five stages from unaware to most aware; a line that works at one fails at another. To an
unaware reader name no product, only their situation.
- Use when / not when: deciding what line 1 may say; not as a template. Whatever the stage,
open on them, never on the product.
- Example: "Spring is when the stations that drifted over winter need a visit."
- Evidence: Practice. Built on his own mail-order results.
### Give-first offer (reciprocity: Regan 1971; Falk, *Econometrica*, 2007; Cochrane review 2009)
- What it says: give something useful the reader keeps, with no strings; the ask after it stays tiny. A
gift sent up front beats one promised on reply; in 10,000 charity letters a gift raised giving 17 to 75%.
- Use when / not when: the first message's ask and follow-up 2, only with an offer the user really made;
never a demo, call or discount dressed as a gift, never one that needs their data.
- Example: "Shall I send you the checklist two teams use to pick which stations to visit first?"
- Evidence: Measured (field trials). Offer ask +28% in email data: Vendor claim.
### Social proof (Goldstein, Cialdini and Griskevicius 2008; backfire: Cialdini et al. 2006)
- What it says: people follow what similar others do; telling them many do the bad thing makes it worse.
- Use when / not when: follow-up 1, one real customer like them, in the user's words; never a crowd, the
seller's totals, a site figure, or a normalised problem ("most managers never check this").
- Example: "A utility with 80 stations now visits only the ones that drift, since they plan with us."
- Evidence: Measured, mixed. The best supported of Cialdini's levers; field replications vary.
### Poke the bear (Josh Braun, "Ditch the pitch, poke the bear")
- What it says: one neutral question about the cost of the current way, never a leading one.
- Use when / not when: follow-up 1 when no real customer result exists; not as the first message's ask
(asking about their problem scored 29% lower there, vendor email data).
- Example: "Do you see a station drift before the monthly data comes in?"
- Evidence: Practice.
### Detach (Josh Braun; Chris Voss, *Never Split the Difference*, 2016)
- What it says: no chasing, no guilt; the last message makes no easy and closes the loop.
- Use when / not when: follow-up 2 and the optional breakup; never "just checking in", never guilt.
Two sentences: the offer, then the question with the way out.
- Example: "I can prepare the checklist for your spring rounds. Shall I send it, or would you rather I
leave it there?"
- Evidence: Vendor claim: breakup language lifted late email replies (observational).
### FAB and the "So what?" test (Neil Rackham, *SPIN Selling*, 1988)
- What it says: a feature, what it does, what it does for this reader; ask "so what?" after each claim
until the answer is their week. Only benefits tied to a need the buyer stated went with wins.
- Use when / not when: a reply to someone who asked or objected, tied to their words; never cold.
- Example: "Unlike a monthly round, each station flags its own drift, so the second trips you mentioned
mostly stop."
- Evidence: Vendor claim (Huthwaite's own observed calls, not independently replicated).
### Yes/no asks (Chris Voss 2016; Josh Braun; Gong Labs)
- What it says: one easy question answered in a word, its object named; never "why", never two questions.
- Use when / not when: every ask; a no-oriented form ("Would it be a bad idea if...?") only as a
follow-up 2 test arm.
- Example: "Shall I send the checklist?"
- Evidence: Vendor claim: interest asks about twice time asks (observational, email).
## Further reading
- Eugene Schwartz, *Breakthrough Advertising*, 1966.
- Becc Holland, "How to personalize at scale". https://www.persistiq.com/how-to-personalize-at-scale/
- Armin Falk, "Gift exchange in the field", *Econometrica*, 2007. https://www.econometricsociety.org/publications/econometrica/2007/09/01/gift-exchange-field
- Robert Cialdini, *Influence*, 1984; Goldstein, Cialdini and Griskevicius, 2008.
- Josh Braun, "Ditch the pitch, poke the bear". https://joshbraun.com/ditch-the-pitch-poke-the-bear/
- Chris Voss, *Never Split the Difference*, 2016.
- Neil Rackham, *SPIN Selling*, 1988.
- O'Keefe and Hoeken, "Message design choices don't make much difference to persuasiveness", *Frontiers in Psychology*, 2021. https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2021.664160/full
- 30MPC, Gong and Outbound Squad, *The Ultimate Cold Email Data Report*, 2025 (chiffre d'éditeur).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |