marketing-may

/marketing-may

Agent skill for doing data-driven marketing.

Run it on your project without additional prompting to get a review of the marketing posture of your project with actionable next steps. See the examples below for more advanced usage.

Install

npx skills add bobisme/marketing-may

How to Use

The skill auto-loads when your prompt mentions marketing, positioning, messaging, copy, landing pages, ads, pricing, channels, funnels, retention, experiments, customer discovery, competitive research, or growth decisions.

Beautiful PDF reports

Ask for a report and you get one — a typeset, editorial PDF, not a wall of Markdown. When you request a report, PDF, or shareable deliverable (a teardown, pricing review, or competitive analysis as a document), the skill renders it with a bundled Typst template: paper-white editorial layout, real typefaces, hairline tables, the critical-unknowns ledger, kill/pivot rules, and appendices — the skill's own analysis structure, typeset. No Typst install required: the renderer fetches a pinned static binary on first use, and nothing leaves your machine (reports are confidential).

Use /marketing-may — produce a marketing & copy teardown of https://your-site.com as a PDF report.

Example output — the bundled sample, docs/example-report.pdf (cover and TL;DR/contents pages):

Or render one yourself from structured content — see the authoring API in typst/:

typst/bin/render your-report.typ out.pdf

Example prompts

The skill routes by intent. Below are prompts that exercise each branch — copy, adapt, or paste your own context.

Add "Use /marketing-may" (claude) or "Use $marketing-may" (codex) to these examples to ensure the skill gets triggered.

Segmentation, ICP & JTBD

• "Help me pick a segment to focus on. We're a Postgres slow-query tool selling to backend engineers, marketing managers, and DBAs — we can't tell which buyer to commit to."
• "Score these 4 segments for us: solo founders, seed-stage CTOs, Series A platform teams, enterprise DBAs."
• "Write an ICP and JTBD map for a hosted Temporal alternative."

Positioning & messaging copy rewrite

• "Rewrite my hero section. Current copy: ''. We sell observability for ML inference."
• "Our landing page isn't converting. Here's the URL / here's the copy — diagnose and rewrite."
• "Give me a positioning statement. Category, buyer, alternative, differentiation, proof, sales motion."
• "Mine this VOC dump for buyer language" (paste reviews / interview notes / support tickets — runs message_miner.py).
• "Lint this landing-page copy" (runs copy_lint.py against vague adjectives, unsupported superlatives, dark patterns, hidden price, missing CTA).

Pricing & offer

• "What should I charge for a self-serve Postgres slow-query tool?" (will refuse a number without WTP evidence; returns a test design instead).
• "Here's a Van Westendorp CSV from 42 respondents — what does it say?" (runs pricing_research_analyzer.py).
• "Design a pricing test. We're considering $49 / $99 / $199 monthly."
• "Critique our packaging — Free / Pro $29 / Team $99 / Enterprise call us."

Competitive intel

• "Are competitors eating us? Direct rivals: Datadog DBM, pganalyze. Substitutes: RDS Performance Insights, manual EXPLAIN."
• "Tear down pganalyze, Datadog DBM, and RDS Performance Insights — features, price, ICP, positioning, gaps."
• "Build a competitor matrix from this CSV" (runs competitor_matrix.py).
• "Find the whitespace in the Postgres tooling category."

Funnel, activation & retention

• "Why isn't our funnel converting? Here's signup → activated → paid for last 90 days."
• "Analyze this funnel CSV by segment" (runs funnel_analyzer.py).
• "Are users sticking around? Here's cohort × week event data" (runs retention_analyzer.py for plateau/declining/flattening verdict).
• "Pick our activation event. We're a usage-based API product."
• "Design an event taxonomy that survives a rewrite."

Metric trees & north-star

• "What should our north-star metric be? We're a two-sided marketplace for dog walkers."
• "Build a metric tree for a sales-led B2B SaaS at $40k ACV."
• "Give me the starter metric tree for OSS-led commercial."

Experiments & stats

• "Should we run this test? Variant A vs. variant B on the pricing page, ~3,800 weekly visitors." (will refuse frequentist at low traffic and route to Bayesian).
• "How big a sample do I need to detect a 2pp lift on a 6% baseline?" (runs stats_cli.py sample-size).
• "We just shipped a test — here's the result. 412 / 9,103 vs. 478 / 9,088. Read it." (runs stats_cli.py srm + bayes).
• "Rank these 12 experiments by Expected Learning Value" (runs experiment_prioritizer.py).
• "Write an experiment plan with pre-registered decision rules."

Channels & outbound

• "Pick a channel for us. ACV $1.2k, 0 brand, 1 founder, no budget."
• "Draft a cold-outbound playbook for platform engineers at 200–2,000-person companies."
• "Score this account list against our ICP" (runs outbound_list_scorer.py for A/B/C tiers).
• "Should I sponsor this newsletter? 38k subs, $4,200, dev audience." (uses 2026 CPM bands).
• "How do I get cited by ChatGPT / Perplexity for queries in my category?" (loads modern_channels.md AEO/GEO patterns).

Marketplaces

• "We're building a marketplace for X. Where do we start?" (loads marketplace_network.md cold-start playbook).
• "What's the liquidity target we should be hitting at week 12?"
• "We have supply but no demand — what's the next move?"

Kill / pivot / narrow

• "Are we kidding ourselves? Here's 6 months of data: 14 paying customers, $890 MRR, 11% monthly churn, CAC $740."
• "Should we kill this product? Should we narrow? Should we pivot? Here's the evidence."
• "Pre-register a kill rule for the next 8 weeks."

Customer discovery

• "I want to run customer interviews. Write me a switch-interview script."
• "Here are 5 interview transcripts — extract pain, outcome, objection, trigger, alternative, proof."
• "Design a commitment menu for ICP-validation calls."

"I have no idea where to start" (vague growth questions)

• "Help me grow this thing."
• "How do I market this?" → triggers workshop mode: runs 00_product_intake interactively, asks one question at a time, refuses to draft assets until ICP and dominant alternative are named.

Worked examples

• "Show me a worked example for an indie devtool getting to revenue."
• "Show me the marketplace cold-start case study."

Response modes

Phrasing shifts the response shape:

• Quick mode (default, ≤400 words) — short question, no data. Always opens with a "Critical unknowns" table, then a one-paragraph diagnosis, smallest artifact, ≤5 next actions, one kill rule.
• Deep mode — paste data, ask for a plan, or name a timeline ("here's our funnel CSV — give me a 30-day plan").
• Workshop mode — paste an intake, or ask "where do we start?" — agent runs the intake template interactively.

Decision-grade questions (pricing commits, channel scaling, kill/pivot, hiring) additionally emit a JSON evidence ledger with kill rules per claim — regardless of mode.

Features

Decision frameworks (references)

• Segmentation, ICP & JTBD — qualification ladder, segment scoring, buyer-psychology / objection model.
• Positioning & messaging — category-strategy decision tree, positioning system, message hierarchy, hero-formula bank, copy quality checklist.
• Pricing & offer design — pricing-model decision table, value-metric fit score, WTP evidence ladder, risk-reversal menu.
• Competitive intel — research table, gap taxonomy, four-group competitor set (direct / adjacent / substitutes / aspirational).
• Channels & cold outbound — sales-motion decision tree, ACV/channel fit heuristic, channel-fit scoring, outbound thresholds and template.
• Modern channels — AI search / AEO / GEO citation patterns, podcast sponsorship math (2026 CPMs), community playbooks for Discord/Slack/Reddit, newsletter sponsorships.
• Marketplace & network effects — cold-start playbook, liquidity targets by marketplace type, take-rate ranges, disintermediation defenses.
• Customer discovery — interview rules, structure, commitment menu.
• Experiments — experiment types, ELV scoring, kill/pivot/narrow rules, decision-rule pre-registration.
• Funnel analytics — stage model, activation event selection, event taxonomy rules, attribution caveats.
• Metric trees — pre-built starter trees for self-serve SaaS, usage-based / API, marketplace, OSS-led, course/community, sales-led B2B, e-commerce.
• Stats primer — when to A/B test, when to refuse, sample-size sanity, SRM as precondition, Bayesian small-traffic guidance.
• Browsing recipe — per-page extraction shapes for competitor homepages, pricing pages, reviews, search/AI-search citations, with stopping rules.
• Compliance & ethics — never-recommend list, substantiation, dark-pattern avoidance.
• Decision trees — "what next?", "which test?", "should we scale this channel?".
• Algorithms & practice digest — evidence tensor, segment attractiveness, offer-market fit, category gravity, objection entropy, proof debt, copy-claim risk filter; field rules in compact form.

Templates

Fillable templates for product intake, customer-discovery interviews, ICP/JTBD/segment maps, competitive teardowns, positioning + messaging + copy, pricing/packaging/offer, funnel instrumentation, experiment plans, channel strategy, kill/pivot/narrow decisions, and asset specs.

JSON schemas

Validatable schemas for competitor teardowns, event taxonomies, experiment plans, marketing assets, pricing research, and an evidence ledger for tracking each load-bearing claim's source, behavioral strength, and kill rule.

Scripts (Python, no third-party deps unless noted)

• competitor_matrix.py — summarize competitor research CSV into category norms, gap counts, and whitespace prompts.
• experiment_prioritizer.py — rank experiments by Expected Learning Value with a decision hint per row.
• funnel_analyzer.py — ordered funnel conversion + drop-off, optional segment grouping.
• retention_analyzer.py — cohort × week-N retention table with directional curve-shape verdict (plateau / flattening / declining).
• message_miner.py — extract repeated buyer language from voice-of-customer text; bucket by pain / outcome / objection / trigger / alternative / proof.
• pricing_research_analyzer.py — Van Westendorp and Gabor-Granger directional analysis with segment cuts.
• outbound_list_scorer.py — score account CSV against weighted ICP criteria (firmographic / tech / role / recency-bounded triggers / negatives) and rank into A/B/C tiers.
• stats_cli.py — A/B test sample size, sample-ratio mismatch (Microsoft p<0.0005 threshold), Bayesian Beta-Binomial P(B>A) with credible interval and expected-loss decision aid.
• copy_lint.py — mechanical check of marketing copy for vague adjectives, unsupported superlatives, dark-pattern urgency, hidden price, weak social proof, missing CTA, buzzword density. Supports inline <!-- lint:ignore <rule> -->.
• validate_artifact.py — validate any JSON artifact against the corresponding schema; auto-detects schema by filename.

Worked case studies

Two filled walk-throughs (synthetic but realistic) showing the full skill flow from intake through decision: an indie devtool reaching $3.4k MRR, and a two-sided marketplace narrowing to first 50 transactions.

Quality layer

• 83-test pytest suite covering every script (sample-size formula, SRM detection, Bayesian determinism, funnel ordering invariant, Van Westendorp edge cases, copy-lint rules, retention curve helpers, outbound recency windows).
• Pyright-friendly with pyrightconfig.json.
• just test / just test-quiet / just setup-dev recipes.