lead-magnet
Generate lead magnet concepts grounded in live competitor research, then BUILD the chosen one, including working interactive apps (quiz, calculator, instant quote, AI-graded audit) with a capture gate, as well as written guides and checklists. Covers info products, SaaS, agencies and local service businesses (price calculators, repair-or-replace checkers, seasonal checklists, financing estimates), plus an automated free-report Zap. Use when someone wants more leads or subscribers, needs top-of-funnel ideas, or asks what to give away for free.
How do I install this agent skill?
npx skills add https://github.com/boringmarketer/boring-marketing-skills --skill lead-magnetIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The lead-magnet skill researches competitor sites to generate marketing assets like quizzes, calculators, and automated reports. It produces functional web applications and automation prompts based on user input and research data. The analysis confirms these operations are standard for its marketing and development purposes, with no security risks detected.
- Socketpass
No alerts
- Snykwarn
Risk: MEDIUM · 1 issue
What does this agent skill do?
Lead Magnet
The best lead magnet is the thing a prospect wants right now, whose use naturally creates desire for the paid offer. This skill finds that thing with live competitive research, pressure-tests it against five concept tests, and then builds it. When the format is interactive, "builds it" means a working app, because an agent can now ship a quiz or a calculator as fast as a PDF. The method is a floor, not a ceiling. When the offer in front of you argues for a better path, take it and say why.
Context
Only if ./brand/, ./brands/ or MARKETING.md exists (nothing to load otherwise), run:
python3 ${CLAUDE_SKILL_DIR}/../start-here/scripts/brand_context.py --for lead-magnet
(${CLAUDE_SKILL_DIR} is this skill's folder; outside Claude Code use the folder this
file was read from, e.g. .agents/skills/lead-magnet. If the script or python3 is
unavailable, read the brand folder directly: ./brands/<slug>/ when ./brands/active
names one, else ./brand/; then MARKETING.md, preferences.md and the files in
reads:. If none exist, proceed standalone and ask only for what the task strictly requires.)
If voc.md exists, it names the magnet topic better than any brainstorm: the most
frequent, most intense problem-aware pain is usually the concept, and its exact phrasing
is the hook. Quote it, don't paraphrase it.
If ./brand/assets.md or ./campaigns/*/brief.md already reference a lead magnet, say so
in one line and ask: revise it, build a new one for a different segment, or generate fresh
concepts.
Modes
- Full run (default): Phases 1-4. Research, concepts, the user picks, build.
- Build-direct / Quick Mode: the user knows what they want ("build me a pricing calculator", "make the quiz from concept 2"). Skip research and concept ceremony, confirm the 3-4 inputs the build strictly needs, go straight to Phase 4.
- Concepts-only: the user wants ideas to react to. Stop after Phase 3.
Phase 1. Understand the offer
Three facts gate everything downstream. Pull them from brand memory; ask only for what's missing.
- Business type. It sets the default format pool and the reference to load:
info product or creator (
references/info-product-magnets.md), SaaS (references/saas-magnets.md), agency, consultant or freelancer (references/services-magnets.md), or a local service business customers call from a map (references/local-magnets.md, and read the Local section below). - The transformation, not the product. What changes in the customer's life after buying. The magnet delivers a small version of that same transformation; a magnet about a different one attracts the wrong list.
- The audience and its AI fluency. If this audience builds with AI, automates with
Zapier, Make or n8n, or buys prompts and agents, load
references/ai-era-magnets.md. Prompt packs, skill packs, workflow bundles, custom agents and AI-graded audits are what they trade in, and that changes the concept pool.
Phase 2. Competitive research
What competitors give away shows the saturated formats; what they pay to promote shows what converts. Get both with parallel subagents, never from memory.
Identify 3-5 direct competitors (from competitors.md, positioning context, or a quick
search of the niche; for local, the businesses that show up for "[service] [city]"). Run
one WebSearch in the main thread first; if it's denied, write a draft from model knowledge
instead of fanning out and say so once in the note. Then, all in parallel:
- One subagent per competitor (3-5): read their homepage, blog sidebars and popups,
/free*,/resources,/tools, the footer, and for local businesses any "price", "estimate" or "financing" page. WebFetch and WebSearch work with no key. We use Firecrawl to read these pages in full, JavaScript-built ones included (free tier of 1,000 credits a month,FIRECRAWL_API_KEY), and Exa to find competitors and resource pages a keyword search misses (free tier, then pay as you go,EXA_API_KEY). Return an evidence pack: every opt-in offer found, with the verbatim hook, format, what's gated and what's free, and the URL. Save the packs to./campaigns/{name}/notes.md; they are working data, not chat. - One subagent for the ad libraries: search facebook.com/ads/library (and adstransparency.google.com) for each competitor plus 1-2 category keywords. Magnets running as paid ads are the strongest conversion signal available; nobody keeps paying to promote a freebie that doesn't convert. Return the advertised magnets, ad hooks verbatim, format, how long the ad appears to have run, and URLs.
Synthesize a short landscape: each competitor's magnet, hook and format; what's paid-promoted (the proven converters); what's saturated (3 of 4 have a PDF guide, so don't ship a fourth); and the open gap (no interactive tool, no one targets segment X, every hook is generic). Show it in a few lines. Never present model knowledge as live research: the note says what it rests on ("Read 6 competitor pages and 4 ads live on Nov 13", or "Competitor read is from model knowledge; verify before spending"), naming any subagent that came back empty.
Phase 3. Concept generation
Generate 3-5 distinct concepts, one starred. Distinct means different format families or different bridge mechanics, not five titles for the same PDF.
The five tests
Every concept passes all five before it's presented. Run them silently: one that fails gets fixed or cut, not padded with rationale, and the client sees no test grid.
- Specificity. Specific outcome, audience, timeframe, named method. Could a competitor use the exact same title? If yes, sharpen it.
- Bridge. Using the magnet creates logical desire for the paid offer. The magnet is step 1 of what you sell; the paid thing is more, deeper, or done for you. If you have to explain the connection, there isn't one.
- Quick win. It solves one problem completely, with a result in the first sitting. A magnet that needs weeks of study is homework, not a gift.
- Value equation (Hormozi): Value = (Dream Outcome x Perceived Likelihood) / (Time Delay x Effort). Most concepts fail on the bottom: too slow to pay off, too much effort to use.
- Defensibility. Can an AI assistant replicate or summarize this away? A gated PDF of known advice gets summarized in seconds. Durable magnets are personalized or interactive (their data, their result: quizzes, calculators, audits, instant quotes), genuinely scarce (original data, benchmarks, community access), or working tools (apps, skills, workflows). Pure information survives only when it's original. Default: 2+ concepts defensible by interactivity or scarcity, and the starred pick almost always is.
Format selection
Interactive formats (quizzes, calculators, audits) are widely reported to convert at several times the opt-in rate of static PDFs. The circulating figures are illustrative, not measured, so trust the ordering, not the numbers. BUILD MODE makes a working quiz no more expensive than a guide, so default to interactive unless the audience or the bridge argues otherwise.
| Format | Reach for it when | Defensibility |
|---|---|---|
| Quiz / assessment | segmentation or self-discovery drives the offer | high (their answers, their profile) |
| Calculator / instant quote | the value is a number (price, ROI, savings, monthly payment) | high (their inputs, their result) |
| AI-graded audit or report | diagnosis comes before the sale (agencies, SaaS, consultants) | highest (their URL or data plus your rubric) |
| Checklist / template | immediate use, a method on display | low to medium; needs an original method |
| Guide / framework | an authority play on a complex topic | low; must carry original data or a point of view |
| Swipe file | curation and annotation are the value | medium; the annotations are the moat |
| Challenge (multi-day) | transformation plus community offers | medium; the cohort is the scarcity |
| AI-era formats | the audience builds or buys with AI | see references/ai-era-magnets.md |
For hooks and conversion psychology (reciprocity, curiosity gap, commitment ladder,
identity), load references/psychology.md and use its seven triggers and its closing checklist as
the bar for concept copy, applied silently.
Seven hook shapes cover nearly every magnet: the Shortcut ("X without the usual pain"), the Secret ("the hidden thing behind the result"), the System ("the NAMED method for the outcome"), the Number ("7 things that X"), the Assessment ("find your type or score"), the Transformation ("from painful state to desired state"), and the Case Study ("how [neutral example] got [result]"). When voc.md holds a repeated verbatim phrase, the hook is usually that phrase.
Present the concepts
Lead with the Quick Pick block (the starred concept, one line). Then for each: name · hook headline · format · bridge (one sentence: the magnet delivers X, which creates desire for the paid offer) · what gets built · best for (channel or situation). The star's rationale names the gap it exploits, the paid-promoted competitor format it improves on, and why AI can't replace it. Do not print per-concept test results. Close with: pick one to build, tweak one, or ask for more.
Calibration example. Say the offer is a $497 course teaching freelance designers to price by value; research found three competitor PDF guides, one of them paid-promoted, and no interactive tools:
Quick Pick: ★ Concept 1: The Underpricing Calculator The niche's proven converter is a pricing PDF running as an ad. Same pain, but a calculator uses their own numbers, so it converts better and an AI summary can't copy it.
1. The Underpricing Calculator ★ "Find out how much your hourly rate is costing you, in 90 seconds" Format: calculator (working app). Bridge: their own numbers show the gap between hourly and value pricing, and the course closes it. Built: the app and the email gate. Best for: paid social and bio link.
2. The Pricing Confidence Quiz. "What's your pricing personality?" Quiz. Bridge: the profile names the habit keeping them cheap and maps to course modules. Best for organic social and segmented follow-up.
3. The Proposal Teardown Swipe File. "11 real proposals that closed at 3x the market rate, annotated line by line." Bridge: shows the artifacts the course teaches. Holds up only because the proposals and notes are real. Best for newsletter swaps.
That's the bar. Each concept names its gap and its bridge honestly, and 2 and 3 are not restatements of 1. A SaaS version of the same bar: an invoicing tool's "Late Payment Cost Calculator" (their invoices and terms in, the cash tied up out) beats a "10 tips for getting paid" PDF for the same reasons.
Local service businesses
For plumbers, roofers, HVAC, dentists, med spas, cleaners and other businesses customers
call from a map, load references/local-magnets.md. What changes:
- The goal is a booked job, not a list. The magnet is a reason to hand over a phone number while they're deciding. Capture phone and zip (the zip checks the service area); email is optional.
- It answers the question asked right before buying: what will this cost, repair or replace, is it urgent, can I afford it. The four that work are an instant quote or price calculator, a repair-or-replace checker, a seasonal checklist with a reminder opt-in, and a financing estimate.
- Every number comes from the owner. Price ranges from their own invoices, thresholds from their own rule of thumb, monthly payments from the lender's real terms.
- Show the price range ungated by default and ask for the phone to book or lock a written quote. Hiding a range behind a form contradicts the price clarity it promises.
- Follow-up speed beats nurture. Someone who just priced a water heater is buying this
week. Route the lead to
/speed-to-leadfor the text-back and call, then/email-sequencesfor the quote follow-up.
Sample (Hartline Plumbing, Dallas TX, a sample business): competitors offer "free estimates" and a tips blog, and none shows a price. ★ The Dallas Water Heater Price Check: "What will a new water heater cost in your home? 30 seconds, no phone call." A calculator with an ungated range and a phone field to book a same-day slot. The reference has the other two concepts and the build rules.
Agencies serving local businesses have a strong magnet of their own, a free
automated report on the owner's website. The worked example in references/report-zap.md
is a four-step Zap (new form entry or sheet row, read the prospect's site, AI writes a
short report as HTML, email it), with the prompt and the lessons from building one.
Phase 4. BUILD MODE
The user picked. Build the thing itself, not a spec or notes for a quiz tool. Confirm the selection in one line, gather only the inputs the build requires (offer details, pricing, real numbers for formulas, audit criteria; never invent these), then produce.
Interactive formats: quiz, calculator, instant quote, AI audit
Load references/interactive-build.md and follow it. The deliverable is a working
app: single-file HTML by default (or a Next.js component if the user has a project to
put it in), with real scoring or pricing logic, 3-5 reachable result profiles where
relevant, the capture gate, and the ESP or webhook stub with the key location commented.
For local calculators, the capture asks for phone and zip and the zip is checked against
the service area. The spec's pre-ship QA is mandatory: open it, click every path, check
the math against a hand-worked case. An interactive magnet that errors on the first click
converts worse than no magnet.
Files: ./campaigns/{name}/app/ (the app and a README) plus
./campaigns/{name}/lead-magnet.md documenting the questions, formulas, price sources and
profiles, and the bridge logic, so /email-sequences and /direct-response-copy can read the
magnet without parsing JavaScript.
Automated report magnets
When the magnet is "your site or data in, a report out by email", load
references/report-zap.md. Build the prompt and the step-by-step setup, run it once on a
real input before handing it over, and save the prompt as ./campaigns/{name}/prompt.md.
AI-era formats: prompt packs, skill packs, workflow bundles, agents, datasets
Load references/ai-era-magnets.md for the bridge logic, then build the actual artifact:
the tested prompts with use notes, the SKILL.md and references of a skill pack, the
workflow JSON and setup README, or the agent's full system prompt and gate page.
Everything executable gets the same bar as interactive formats: test it before shipping.
Written formats: checklist, template, guide, swipe file, challenge
- Checklist: 10-25 items grouped by phase; each is an action, why it matters, and a gotcha; a quick-start top 3; a bridge section and CTA at the end. For a seasonal local checklist, split it into "you can do this" and "call a pro".
- Template: instructions, [BRACKETED] fill-ins, guidance per section, one fully filled example, the bridge.
- Guide: a TL;DR up top; 3-7 sections, each principle, why, how, example; a "put it together" section; the bridge. 1,500-3,000 words.
- Swipe file: 20-50 items by category, each with its source and why it works. The annotation is the defensibility, so it can't be thin.
- Challenge: day by day, each with a 15-30 minute action, a teaching point, a success metric, an engagement prompt and a subject line; the final day bridges. The emails themselves belong to /email-sequences.
Every written magnet passes the Phase 3 defensibility test as built: if the draft reads like something an AI assistant would produce from the title alone, it isn't done. Add the user's data, examples, opinions and numbers until it couldn't have come from anyone else.
Quality gate
Before presenting, check the magnet against its own promises, silently: the quick win is achievable
in the stated time; the bridge reads as a natural next step, not a pitch; the voice matches
voice-profile#tone-spectrum (read result-profile and bridge copy against it); interactive
apps passed the QA walk; no fabricated statistics anywhere, and every number is the
user's real number or clearly marked illustrative. Fix what fails, then present only the
passing magnet. If learnings.md holds a prior (a hook or format that performed) that
changed the build, give it one line in the note; otherwise say nothing about it.
Output
Files first, chat second.
./campaigns/{magnet-name}/
lead-magnet.md ← full content, or the documented logic for app formats
app/ ← interactive formats only: index.html (or component + API route), README.md
prompt.md ← automated report magnets only: the full prompt and the Zap steps
brief.md ← goal, format, hook, audience, bridge, paid offer,
differentiation vs researched competitors, distribution plan
notes.md ← working file: competitor evidence packs, ad hooks, URLs, QA results
Magnet name: lowercase-kebab from the concept ("The Dallas Water Heater Price Check" →
water-heater-price-check). lead-magnet.md carries frontmatter: title, format, hook,
bridge_to, target_audience, consumption time, status, date. Append one line per asset to
./brand/assets.md.
Chat delivery per start-here/references/output-format.md: ## LEAD MAGNET: {name} plus
one line of date and what it is for, a content summary (what's in it, the bridge in one
line, and for apps how to deploy and where the ESP key goes), one short Sources and
assumptions note (10 lines max, including any real numbers the owner still has to
supply), Files saved, What's next. The QA walk stays in notes.md.
When the user edits your work
Apply the edit, then make it stick: save a gold case to ./brand/gold/<this skill>/ and
one to three yes/no checks to ./brand/checks.md, and confirm in one line. Before handing
over any work, answer the checks the context block loaded, silently, and fix every no.
How: ${CLAUDE_SKILL_DIR}/../start-here/references/taste-loop.md.
What's next
/landing-pagefor the opt-in page, built on this magnet's hook and bridge (~20 min)/email-sequencesfor delivery and follow-up; for quizzes and audits, per-segment sequences keyed to the result profiles; for local, the quote follow-up (~15 min)/speed-to-leadfor local magnets, so every price check or quote request gets a fast call or text (~15 min)/content-atomizeror/paid-adsto drive people to it, using the magnet's hook (~15 min)
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/boringmarketer/boring-marketing-skills/lead-magnet">View lead-magnet on skillZs</a>