The executable-product category is real, but the observed price ceiling is ~$200 — MAC's $350 would sit above every verified comparable
The question
What are buyers actually paying for executable / installable / agentic info-products in 2026 — concrete pricing, top-of-funnel mechanics, and refund rates for 5-10 named comparables? Context: MAC v0 is mid-build and the founder is gating launch on whether the "executable course" category is a real market or just a coined term.
What we already know (from the vault)
- MAC's price is already decided at $350 one-time, not a $300-2k band. [[2026-05-14-mac-pricing-intent]] logs the founder-confirmed decision (iMessage 2026-05-14 15:41 ET) plus three locked positioning principles: no-brainer price, sold INTO the internal data engineer, and an explicit acknowledgement that "multi-thousand-dollar value exists but isn't extractable at this price" without a high-touch sales motion. The Research Backlog note's "$300-2k guess" framing is stale relative to the vault.
- The vault's refund-rate model for MAC is 8-15%, derived from traditional-course data, not from executable-product data. [[2026-05-18-low-ticket-dev-tool-launch-friction]] documents the Q1 2026 environment: platform-wide info-product refunds ~21%, the $497-$1,997 premium-course band ~28%, paid challenges <4% (Communipass). The single strongest A/B is the Aamir case: $497 self-paced course → 412 enrollments, 22% refunds, 6% completion; reframed to a $247 paid challenge → 184 enrollments, 3.2% refunds, 77% completion. The brief's own caveat: no public launches were found in the $150-$500 developer-tool segment with full refund numbers.
- The agent-distribution market is a TOOL market, and almost nobody charges a take-rate. [[2026-06-20-bookstores-for-agents-2026-live-comps]] catalogues eight marketplaces: Claude Skills (free to publish, no paid tier), mcp.so / Smithery / PulseMCP (free, curation-only), Hugging Face Spaces, LangChain Hub, Vercel Agent Gallery — all free. The paid tier competes on take-rate: MCPize 85% creator share, Apify ~80% minus compute, Vinkius (launched Mar 25 2026) zero commission / compute-fee only. GPT Store pays an opaque usage-weighted share; Replit Agent Market takes an app-store-style cut on builder-priced sales.
- The vault's own conversion evidence says bounded, self-evaluable skill outputs are what get paid for. The 368-probe field study cited in [[2026-06-20-bookstores-for-agents-2026-live-comps]] found skills/eval endpoints convert ~33% while live data feeds convert 0-0.8% at identical price. That is the strongest empirical support for the executable framing — it just isn't retail-buyer data.
- Payment rail and distribution shape are already settled. [[2026-05-18-mac-platform-comparison-stripe-vs-ls-vs-gumroad]] lands on Lemon Squeezy (~5% vs Gumroad's ~13%, native license-key delivery) and [[2026-05-04-mac-product-shape-decisions]] records the Stripe → GitHub-token install flow and the explicit intent to coin the executable-course category.
What the web says
- CLSkills Hub (clskillshub.com) is the closest live full-catalog comparable, and its prices are far below MAC's. Verified from its pricing page: The Cheat Sheet $15 one-time (120 tested prompt codes, 30 days email support); Cheat Sheet + Skills Library bundle $30; Skills Library Lifetime $25 (845 verified skills); Realtor Skill Pack $19 launch pricing, $49 regular (25 ready-to-use Claude assistants for US real estate agents, Fair Housing appendix, 5-minute installation); Legal Pack $200 (10 flagship skills today, 40 more rolling out through Q3 2026, lifetime updates); Team Pack (5 people) $199. Free tier: The Claude Guide, Prompt Stacker, 20 Free Agents, and 1,545 Community Skills (clskillshub.com/pricing).
- No refund or guarantee language appears anywhere on the CLSkills pricing page, and no install or customer counts are stated. This is the norm, not an outlier — see the contradictions section.
- Agensi is the only marketplace publishing seller-economics data, and it is brutal. Agensi runs a curated SKILL.md marketplace: 30% platform fee, creators keep 70%, Stripe Connect payouts, curl one-liner install to
~/.claude/skills/, 8-point security scan before listing. Observed pricing: individual skills $5-$25 one-time, with some subscription models emerging. Earnings: top-earning individual skills generate roughly $500-$3,000/month; the median listed skill earns less than $50/month; most revenue concentrates in the top 10% of listings. Figures described as drawn from publicly available Q1-Q2 2026 marketplace data (agensi.io). - A single named unit-level data point: an Agensi keyword-research skill priced at $7 with 89 installs (≈$623 gross lifetime, ≈$436 to the creator at 70%), cited in Agensi's own make-money guide (agensi.io/learn/how-to-make-money-with-ai-2026) — search-snippet sourced, not independently fetched.
- The largest skill catalog in the ecosystem is free and open-source.
jeremylongshore/claude-code-plugins-plus-skillsadvertises 425 plugins, 2,810 skills, 200 agents distributed through tonsofskills.com with accpiCLI package manager, at no charge (GitHub, tonsofskills.com). Free abundance is the primary competitor to any paid skill product. - explainx.ai runs a free public SKILL.md registry and publishes zero pricing evidence. Its monetization guide offers only positioning heuristics (single utility → lower one-time price; multi-step workflow → mid-tier; regulated/high-risk → premium; team rollout → seat-based) and explicitly warns that "counts, marketplace fees, and third-party payouts change frequently — verify upstream." No observed revenue figures, refund rates, funnel metrics, or take rates are provided (explainx.ai).
- The adjacent n8n workflow-template economy prices the same capability 20-100x higher when sold as a service, not a product. Lead-automation workflow builds run $400-$1,200 for a pilot and $800-$2,000 when CRM rules, dedupe, and multiple sources are involved (browseract, clickwise); n8n's own cloud tiers run €24 / €60 / €800 per month. The $800-$2,000 numbers are done-for-you engagements, not self-serve installable products — do not read them as product-price comparables.
- No refund rate was found for any named executable/installable product. Three fetches and three searches surfaced zero refund, chargeback, or completion data for paid Claude skills, skill packs, Custom GPTs, MCP servers, or Bedrock agents. The category is too young and too small for anyone to be publishing it.
Convergences and contradictions
- Convergence: verticalization is the only observed price lever, and it is worth roughly 10x. CLSkills' horizontal products sit at $15-$30; its vertical packs sit at $49 (Realtor, regular price) and $200 (Legal). Agensi's horizontal skill band is $5-$25. explainx's heuristic independently predicts the same shape ("regulated/high-risk → premium"). This converges cleanly with the vault's [[2026-05-18-low-ticket-dev-tool-launch-friction]] finding that tool-shape and outcome-anchoring — not price — drive buyer behavior.
- Contradiction with the working premise: MAC's $350 is above every verified self-serve comparable in this scan. The highest verified single-price installable product found is CLSkills' $200 Legal Pack; the highest team price is $199. Nothing in the $300-$2,000 band has a live, verifiable, self-serve comparable. The $1,000-$2,000 end of the backlog note's guessed band is entirely unevidenced — every number found near it is a done-for-you services engagement, not a product.
- Contradiction inside the marketplace layer: Agensi's 30% take is the outlier high, and it is the only source with earnings data. MCPize takes 15%, Apify ~20% of compute-adjusted revenue, Vinkius takes 0% ([[2026-06-20-bookstores-for-agents-2026-live-comps]]). The take-rate is racing to zero before there is volume to take a cut of — which means Agensi's published $500-$3,000/mo top-earner figures are simultaneously the best evidence we have and the most self-interested source in the scan.
Synthesis for RDCO
The founder's gate resolves YES on mechanics and NO on price. The category is unambiguously real: multiple independent operators are selling installable, paywalled, one-command-install agent skills through Stripe today, with a working discovery-to-install funnel (marketplace SEO, curl one-liner, security-scan trust signal, agent-native activation where the skill auto-triggers off the task description). CLSkills has shipped a full commercial ladder; Agensi has shipped a curated storefront with published seller economics; MCPize, Apify, and Vinkius are competing on creator share. MAC is not inventing a category out of nothing — the plumbing MAC assumes (Anthropic plugin marketplace + Stripe paywall) is standard practice as of mid-2026. But the price MAC assumes is not. Every verified self-serve installable product in this scan clears at $5-$200, the median Agensi listing earns under $50/month, and the free catalogs are enormous (2,810 skills at tonsofskills, Anthropic's own Skills surface free to publish). MAC at $350 would be the most expensive self-serve installable skill product this scan could find.
That is not automatically fatal, and the reason is the shape of the price distribution, not its ceiling. The observed $5-$30 cluster is prompt-pack-shaped: horizontal, undifferentiated, competing directly with free. The moment a product verticalizes and attaches to a regulated or expensive outcome, price jumps roughly 10x — Realtor $49, Legal $200. MAC is verticalized (data engineering), outcome-anchored (140,937 ghost rows, 5→105 tests in 6 weeks per [[2026-05-14-mac-pricing-intent]]), and sells against a build-vs-buy alternative priced at 8-40 hours of $200-300/hr principal-engineer time. That is a materially stronger value story than a Realtor skill pack, and it justifies sitting at the top of the observed curve. It does not justify sitting 75% above the top of it on faith. The honest read is that $350 is a defensible stretch and $200-$250 is the evidenced landing zone — and the delta is a testable question, not a conviction question. The vault's own revisit condition already anticipates this ("first-100-sale data shows price elasticity").
On refund rates: we do not have them, and we should stop modeling against numbers borrowed from a different category. There is no published refund, chargeback, or completion data for any paid installable skill product. The vault's 21%/28% figures are traditional-course numbers from Communipass, and [[2026-05-18-low-ticket-dev-tool-launch-friction]] already flags that MAC's tool-shape should beat them. Fine — but "should beat them" is an untested hypothesis that MAC's own launch would be the first public test of. The practical consequence: instrument refunds at 14-day and 60-day marks from sale #1, treat the first 100 sales as the category's first refund dataset, and do not put a refund-rate number in any forecast the founder acts on without labelling it unverified. The one adjacent number worth holding is the Aamir A/B's directional claim that moving time-to-first-value to Day 1 cut refunds 7x — which is exactly what MAC's <5-minute unboxing P0 is for.
The strategic risk this scan surfaces is not pricing, it is free. MAC's real competitor is not CLSkills' $200 Legal Pack; it is the 2,810-skill free catalog and Anthropic's own zero-cost publishing surface. Every operator in this scan runs a large free tier as top-of-funnel — CLSkills gives away 1,545 community skills, 20 agents, and a guide before charging $15. RDCO currently has no free MAC surface. If MAC launches as a cold $350 paywall with no free artifact in the ecosystem doing discovery work, it is fighting the category's dominant funnel shape rather than using it. The cheapest version of the fix is to publish one genuinely useful free skill (a scoped /dq audit against a sample project) to the public marketplace as the top of MAC's funnel, and keep the matrix methodology, the test-generation grammar, and the triage runbook behind the paywall.
Why this is in the vault
This is the competitor evidence the founder explicitly gated MAC v0's launch on: it confirms the executable-product category exists commercially (unblocking launch) while contradicting the assumed price band, which makes it directly load-bearing on the [[2026-05-14-mac-pricing-intent]] $350 decision and on whether MAC ships a free top-of-funnel skill before the paywall.
Open follow-ups
- What is CLSkills' actual sales volume on the $200 Legal Pack and $199 Team Pack? Price-list presence is not demand evidence, and these two are the only verified data points anywhere near MAC's band.
- Has any paid installable skill or plugin publicly reported a refund or chargeback rate? If not by Q4 2026, MAC's own first-100-sale data becomes the category benchmark and is worth publishing as positioning.
- Is Agensi's "top skills earn $500-$3,000/mo" claim independently verifiable, or is it marketing for a 30%-take marketplace? The take-rate outlier status makes this the single most self-interested number in the scan.
- What do Custom GPT builders actually net under the GPT Store's usage-weighted revenue share, and has OpenAI published any per-builder payout figures since the store's monetization launch?
- Does a free top-of-funnel skill measurably convert to paid in this ecosystem, or does free-catalog abundance cannibalize the paid tier? CLSkills' free-1,545 → paid-$15 ladder is the natural case study if any conversion data can be obtained.
- Are there paid AWS Bedrock agent blueprints or marketplace listings with published prices? This scan found none and the question named the category explicitly — the absence may itself be a finding.
- At what price does MAC's build-vs-buy math stop resolving toward buy for an IC data engineer expensing it personally vs. putting it on a team card? The $199 CLSkills Team Pack suggests a seat-priced team SKU is the category's answer to that.
Related
- [[2026-05-14-mac-pricing-intent]]
- [[2026-05-18-low-ticket-dev-tool-launch-friction]]
- [[2026-06-20-bookstores-for-agents-2026-live-comps]]
- [[2026-05-18-mac-platform-comparison-stripe-vs-ls-vs-gumroad]]
- [[2026-05-04-mac-product-shape-decisions]]
- [[2026-05-23-agent-deployer-competitor-pricing-scan]]
Sources
- Vault: [[2026-05-14-mac-pricing-intent]] (
~/rdco-vault/01-projects/mac/2026-05-14-mac-pricing-intent.md) - Vault: [[2026-05-18-low-ticket-dev-tool-launch-friction]] (
~/rdco-vault/06-reference/research/2026-05-18-low-ticket-dev-tool-launch-friction.md) - Vault: [[2026-06-20-bookstores-for-agents-2026-live-comps]] (
~/rdco-vault/06-reference/research/2026-06-20-bookstores-for-agents-2026-live-comps.md) - Vault: [[2026-05-18-mac-platform-comparison-stripe-vs-ls-vs-gumroad]] (
~/rdco-vault/06-reference/research/2026-05-18-mac-platform-comparison-stripe-vs-ls-vs-gumroad.md) - Vault: [[2026-05-04-mac-product-shape-decisions]] (
~/rdco-vault/01-projects/mac/2026-05-04-mac-product-shape-decisions.md) - Vault: [[2026-05-23-agent-deployer-competitor-pricing-scan]] (
~/rdco-vault/06-reference/research/2026-05-23-agent-deployer-competitor-pricing-scan.md) - CLSkills Hub — Pricing — fetched 2026-07-23; primary source for all CLSkills price points
- Agensi — How to Monetize SKILL.md Skills as a Developer: 2026 Guide — fetched 2026-07-23; 70/30 split, $5-25 band, $500-3,000/mo top earners, <$50/mo median
- explainx.ai — Monetize AI Skills: Pricing, Distribution, Playbook — fetched 2026-07-23; heuristics only, no observed figures (negative result)
- Agensi — How to Make Money with AI 2026 — search-snippet sourced, not fetched; $7 keyword-research skill / 89 installs
- GitHub — jeremylongshore/claude-code-plugins-plus-skills — 425 plugins / 2,810 skills / 200 agents, free
- tonsofskills.com — Claude Code Skills Hub
- browseract — How to Make Money with n8n Workflow Automation 2026 — $400-1,200 / $800-2,000 service engagements (not product comps)
- ClickWise — How to Make Money with n8n in 2026