agentleFS
aya-talebaya-taleb/biz-boost-skills/skills/referral-asks/SKILL.md

referral-asks

Skill1 starsChanged 23 days agoMITView on GitHub

Finds who could introduce you to the customers you want, from your own people list, customer list or the LinkedIn connections export you paste or download yourself, then writes short, low-pressure ask drafts, including a ready-to-forward blurb so the intro takes them one minute. Use it when someone says "who could introduce me to", "ask for referrals", "warm intros", "how do I ask my customers for referrals", "I know people at the bar association", "here are my LinkedIn connections, who can help", "write a referral ask", or "my first customers came from referrals, how do I get more". Writes sales/referrals.md and adds introduced leads to sales/leads.csv. Drafts only, the user sends. Never scrapes LinkedIn.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When an intro happens, the new lead goes in with the lead-gen tool, source "referral from <name>": python3 …

    line 26
  2. Distinctive:

    The user's own data only: Use lists the user owns or exported.

    line 37
  3. Distinctive:

    Easy to say no: Every ask says plainly that no is fine.

    line 42
  4. Distinctive:

    Thank-yous and incentives honestly: If the user offers a referral reward, say it plainly in the ask.

    line 46
  5. Distinctive:

    Drafts only, run humanize: If humanize is not installed, clean by hand and say so.

    line 49
  6. Distinctive:

    Use names, not pronouns you were not given.

    line 98

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

referral-asks

Warm intros win more often than cold outreach because trust carries over. Most founders do not ask, or they ask in a way that is hard to say yes to ("let me know if you know anyone"). This skill finds the few right people and writes an ask that is specific, easy, and fine to ignore.

What this skill reads and writes

Reads:

  • Business context from founder-context (founder/context.md, else ~/.founder/active and ~/.founder/businesses/<slug>.md): who the user wants to reach (ICP), the offer, tone.
  • The user's people, from what they give you: a pasted list, a customer export, notes, or the LinkedIn connections CSV they download themselves (LinkedIn Settings, "Get a copy of your data", Connections). Productivity's people notes (~/.founder/people/) if they exist.
  • sales/icp.md and sales/leads.csv, to see which target companies a contact might know.

Writes:

  • sales/referrals.md: who to ask, why, the ask drafts and a tracking table.
  • When an intro happens, the new lead goes in with the lead-gen tool, source "referral from <name>": python3 <skills>/lead-gen/scripts/leads.py add sales/leads.csv --business <slug> --company "<company>" --role "<role>" --source "referral from <name>" ... <skills> is the parent of this skill's folder; lead-gen is installed beside it. If lead-gen is missing, edit sales/leads.csv by hand and keep its 16 columns.

Next skill: lead-email or follow-up once an intro email lands (reply on the intro thread); call-prep when the intro turns into a meeting.

Rules

  • The user's own data only. Use lists the user owns or exported. Never log in to LinkedIn, scrape profiles, or guess who knows whom from public posts.
  • Never invent a connection. "Priya might know Dana" needs a reason from the data (same company, same industry group, a past customer of hers). Label each reason as fact (from the list) or guess.
  • Easy to say no. Every ask says plainly that no is fine. Ask each person once, plus one gentle reminder at most.
  • One ask per person, specific. A named company or a clear type ("a small family law firm in Bellevue"), not "anyone who might need this".
  • Thank-yous and incentives honestly. If the user offers a referral reward, say it plainly in the ask. Do not invent one. Some professions limit referral fees (lawyers, for example, under their bar's rules); suggest checking before offering money to them.
  • Drafts only, run humanize. If humanize is not installed, clean by hand and say so.
  • When the request is off limits (logging in, scraping, mass asks): the whole reply is about 6 lines. Two lines saying no and why, two lines on downloading their own connections export, one offer to pick 5 to 10 people from it. No plan, no templates nobody asked for.

Step 1: Who could help

Sort the user's people into:

  1. Happy customers: the strongest. They know the product works. Ask them who else has the same problem.
  2. People inside or close to the target companies: worked there, works there, sits on a board or committee with them, is their supplier.
  3. Connectors in the buyer's world: association organizers, accountants or consultants who serve the same customers, coworking managers, local business groups.
  4. Friends and past colleagues who simply know a lot of people in the area or industry.

Pick 5 to 10, not 50. For each: who, which group, who they could open a door to and why (fact or guess), and how warm the relationship is. Warmth comes only from what the user says ("we're close", "met once"). A connection date or job title says nothing about warmth: write "unknown, tell me" and ask in the reply. Get each person's role at their company right from the data; do not guess who reports to whom.

With a LinkedIn connections export, match Company and Position columns against the ICP and the companies in sales/leads.csv. A match is a starting point, not a promise they know the decision maker.

Step 2: Write the asks

Each ask is under 100 words and has:

  • a personal first line (from what the user tells you about them; if nothing, use {{personal line}}),
  • what the user is doing, in one sentence,
  • the specific ask: "Do you know the office manager at Juniper Family Law, or anyone at a small family law firm on the Eastside?",
  • the easy out: "Totally fine if not, or if it's awkward.",
  • the offer to make it easy: "If yes, I've put a short blurb below you can forward."

The forwardable blurb (50 to 70 words, written so the contact can paste it): who the user is, what they help with, why this person might care, and how to reply. No hype, no claims beyond founder-context.

For happy customers, a different ask: thank them first for something specific, then "who else do you know who deals with <the problem>?" Give them the blurb too.

Adapt the channel: email, text, or LinkedIn message (user sends by hand). For texts, cut to two lines and skip the blurb unless they say yes.

Personal first lines come only from the user's notes. If you have nothing, write {{personal line}}; do not make up a shared memory, and do not add details to how they met. Use names, not pronouns you were not given. A text is two lines; if it needs more, it is an email.

Groups: for a class, a club or a group chat, write one group post (3 to 4 lines, with the blurb and one clear ask), not one message per person. Then pick the 2 or 3 members the user knows best for a personal follow-up.

Partners who see your buyers every day (a food truck, a front desk, a coworking manager): give them something to hand over, not just a question. A card or a sign-up link or QR code ({{sign-up link}}) they can show, plus a line they can say, plus what is in it for them if the user offers something.

Step 3: Track it

A table in sales/referrals.md: person, asked on, for whom, answer, intro made, lead added. Suggest a reminder after one week, once, then let it go.

File format

# Referral asks for <business>, <date>

## Who to ask
| person | group | could open a door to | why (fact or guess) | warmth |

## Asks
### <person> (<channel>)
<draft>

### Forwardable blurb
<blurb>

## Tracking
| person | asked | for | answer | intro | lead added |

In the reply: the top 3 people to ask first and why, one sample ask, and a last line naming the next skill ("Next: when an intro lands, run lead-email or follow-up on that thread."). Count the words before you save: asks under 100, texts two lines, blurbs 50 to 70.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.