Skip to content
Content
Skill

/warm-intro-finder

Finds the human bridges between the writer and a target — mutual contacts, alumni networks, ex-colleagues, podcast connections, board overlaps — and ranks them by how likely they are to make a useful introduction. Different from connection-finder (which finds "like you" content

From plugin
winning-writing
1731 skills

Claude Code

$ npx -y skills add kalyvask/winning-writing --skill warm-intro-finder --agent claude-code

SKILL.md

warm-intro-finder.SKILL.md
name: warm-intro-finder
description: Finds the human bridges between the writer and a target — mutual contacts, alumni networks, ex-colleagues, podcast connections, board overlaps — and ranks them by how likely they are to make a useful introduction. Different from connection-finder (which finds "like you" content hooks). This one finds *people* who can vouch. Use whenever the user is about to send a cold email and hasn't checked who they know in common with the target. Triggers on "warm intro," "find a connection," "who do I know," "mutual contact," "do I have anyone," "who can introduce me."

Warm-intro finder

Source: `points/cold-email-rules.md` rule 4 ("Find a mutual contact and name them") + the AB test result that the same email sent to `jobs@anthropic.com` got zero response while sent direct via warm intro got a reply at 8:25 AM the next morning.

The premise

A warm intro from someone the recipient trusts roughly **doubles** open rate and **triples** response rate compared to the same email sent cold. Most cold emails fail not because they're bad, but because the writer never checked their network.

This skill exists because the check is mechanical and almost everyone skips it.

What "warm" actually means

Not all mutual contacts are useful. The bar is:

  • The connector has **engaged with both parties recently** (within the last 12 months) — a forgotten LinkedIn link from 2014 doesn't count
  • The connector **has positive standing with the recipient** — a fired ex-employee is worse than a stranger
  • The connector is **willing to spend social capital** — being a 1st-degree connection is necessary but not sufficient

Rank candidates by these three filters, not just by connection-degree.

What to search

Cross-reference the recipient (from `recipient-research`) against the user's profile (from `context/about-me.md`) along nine bridge categories. Highest leverage at top.

1. Recent direct collaboration (1st degree)

Anyone the user has shipped a project with or worked closely alongside in the last 24 months who also has a current connection to the target.

2. Investor / board overlap

If the target is a founder or exec, who funds them? Who sits on their board? The user's network of investors and advisors is the highest-leverage bridge to founders.

3. Same-cohort alumni (operative cohort)

Not just "we both went to Stanford" — same year, same program, same section. *"Section H 2026"* opens doors that *"Stanford alum"* doesn't.

4. Ex-colleagues at the recipient's current company

People who have left the recipient's company in the last 24 months and remain on good terms. They know the culture and can frame the email correctly.

5. Podcast / media connectors

If the recipient appeared on a podcast, who else has been on that podcast? Who hosts it? Who runs the producer? These are warm-ish but underused.

6. Mutual investments / advisor pools

Has the user advised or invested in companies the recipient has also touched? Tangential but real.

7. Conference / event co-attendees

Same panel, same dinner, same off-site. Specifically — same *small* event. Big events (TED, Davos) don't qualify.

8. Local geography + activity

Both serious cyclists in Marin. Both regulars at the same coffee shop. Specific shared physical context creates an easy "hey, I'm in town" pretext.

9. Shared mentor or boss

Not the same school — the same *person*. "We both worked under [name] at [company]" is one of the strongest bridges.

Output format

# Warm-intro candidates for: <recipient name>

## Tier 1 — Strong candidates (use these first)
| Connector | Bridge type | Standing with target | Last interaction with user | Likelihood of helping (1-10) | Action |
|---|---|---|---|---|---|
| <Name> | <e.g., investor in target's company; user worked at portfolio co.> | Strong | March 2026 | 9 | Ask for forwarded email with one-line vouch |

## Tier 2 — Possible candidates (worth a check)
[same table format]

## Tier 3 — Weak / risky (skip unless desperate)
[same table format, with a "why risky" column]

## Suggested approach for each Tier 1 candidate
For each:
- The exact ask (forward an email? do a 3-way intro? mention you in their next chat with the target?)
- The 2-sentence message the user should send the connector (NOT the recipient)
- How long to wait before pinging the recipient directly if no answer

## If no candidates exist
Say so honestly. Recommend three places to look that the skill couldn't see:
- Group chats / Slack workspaces (the skill can't see these)
- Recent LinkedIn 2nd-degree connections (the skill needs the user's LinkedIn)
- The recipient's recent podcast guests (often a 1-call-away bridge)

If the user genuinely has no warm path, recommend whether to send cold anyway, get on their podcast first, or pick a different target.

How to feed candidates

The skill needs at least one of:

1. The user's `context/about-me.md` (with company history, schools, advisor relationships) 2. A LinkedIn URL for the user (for 1st/2nd-degree analysis — the skill flags what it would search if it had access) 3. A pasted list of the user's 50 most-recent meaningful interactions (calendar, Slack, email)

If none are provided, ask. Do not fabricate connections.

The Constine forwardable blurb (template)

Once you have a Tier 1 connector willing to forward, send them a clean, short email they can pass through with a one-line vouch on top. Constine's format:

Subject: [Startup name] — [tagline], [traction or momentum highlight], raising [$X]

Hey [Connector],

Thanks for offering to connect us with [target name]. Sounds like they could be a great fit.

As a quick overview, we're building [company name], [one-line description].
[Short background — founder fit, ROI, mission, or big vision in 1–2 sentences].
Deck here: [link].

A few highlights from our recent progress:
- [Build progress / what you've shipped / proof of efficacy]
- [Big customer names, partner
Read more
@kalyvask
Ships withwinning-writing

31 Claude skills for cold outreach, op-eds, pitches, press inquiries, bios, exec memos, performance reviews, spoken-delivery talks, fact-checking, and reply-rate tracking.

Stars on GitHub

Momentum

stars in total
17
Star history of kalyvask/winning-writing on GitHub, a chart by star-history.com

Other skills on winning-writing.

bluf-rewriter

bluf-rewriter

Reorganizes a memo, email, status update, or report so the bottom line is up front. Implements Kramon's BLUF (Bottom Line Up…

compression

compression

Cuts a draft to a target word count without losing substance. Two modes behind one skill. target-count hits a specific number…

fun-angle

fun-angle

Finds the unexpected, dry, or self-deprecating angle that makes an email memorable. Turns "professional and polite" into "this…

More skills from other plugins.

prompt-lookup

prompt-lookup

Activates when the user asks about AI prompts, needs prompt templates, wants to search for prompts, or mentions prompts.chat. Use…

@gh-f@gh-fInstall Skill
ppt-master

ppt-master

AI-driven presentation workflow for generating editable PPTX decks and slides, reconstructing page visuals, creating reusable…

@hugohe3@hugohe3Install Skill
morph-ppt

morph-ppt

Use this skill when the user wants a .pptx with smooth cross-slide animation — PowerPoint Morph transitions, Keynote-style…

officecli11d33k
@iofficeai@iofficeaiInstall Skill

Questions about warm-intro-finder.

Which coding agents can use warm-intro-finder?
warm-intro-finder installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf with the skills CLI, using each agent's flag shown on this page. Flowy's auto-invocation, which fires a skill when your prompt matches, runs in Claude Code only.