---
type: "article"
title: "API Evangelist Weekly — The Skills Became a Supply Chain, and Extending the Surface Area of This Newsletter"
summary: "12,767 third-party posts across 2,795 provider blogs between October 4th and 10th, as I keep widening the set of API provider blogs I read each week. Agent skills became a supply chain, with VirusTotal studying 35,878 of them and an npm SDK for agent infrastructure shipping credential-stealing malware, while Cloudflare became the first big MCP server to serve skills over the protocol. MCP authentication grew real patterns from Okta, Autodesk and MuleSoft. And agents started choosing between payment rails."
newsletter: "API Evangelist"
newsletter_handle: "apievangelist"
newsletter_url: "https://usecommune.com/n/apievangelist"
author: "Kin Lane (@kinlane)"
published: "2026-10-12T14:28:59.005Z"
canonical_url: "https://usecommune.com/n/apievangelist/a/xXV7FMNt"
markdown_url: "https://usecommune.com/n/apievangelist/a/xXV7FMNt.md"
chat_url: "https://usecommune.com/n/apievangelist/a/xXV7FMNt/chat"
body_source: "native"
likes: 0
replies: 0
body_words: 2037
---

# API Evangelist Weekly — The Skills Became a Supply Chain, and Extending the Surface Area of This Newsletter

As the number of APIs I profile for APIs.io continues to expand, so does the surface area of this newsletter. The number of blog posts providers publish shifts from week to week, but I did another refresh of API provider blog sources, which resulted in me reading **12,767 posts from 2,795 provider blogs**, up from 7,816 posts from 1,705 blogs the week before.

There is no way I can read all of these posts, but I can quickly filter them by topic and keyword using [blogs.apievangelist.com](https://blogs.apievangelist.com/), which has essentially become my new feed reader. From there I prioritize the week's themes and trends, and align them with the storytelling I am doing on the API Evangelist and APIs.io blogs. The result is the breakdown in this newsletter each week, capturing the pulse of the API space powering the AI market.

MCP and Agent Skills definitely dominate the conversation, but this process lets me keep my finger on the API signal within all of the AI noise, showing how important APIs have become even as the vocabulary on the street shifts.

![A dark neon observatory where a great circular lens on a pedestal has come into focus, its sharp centre revealing a dense field of lights that blurs into soft points around it.](https://kinlane-images.s3.amazonaws.com/apievangelist/api-evangelist-images/newsletter/2026-10-12-lens-in-focus.png)

## Agent Skills Became a Supply Chain

Three weeks ago I wrote that [Skills over MCP is final, and now it needs servers](https://apievangelist.com/2026/09/22/skills-over-mcp-is-final-and-now-it-needs-servers/). This week one of the biggest arrived: [the Cloudflare API MCP server now serves Cloudflare skills through the Skills over MCP extension](https://developers.cloudflare.com/changelog/post/2026-10-10-cloudflare-mcp-skills/). Clients that support it discover them with `skills/list` and read the files at `skill://` URIs. A skill is no longer a folder you copy into a repo. It is something a server hands an agent at connection time.

![A dark neon loading dock where a long conveyor carries small glowing parcels toward a waiting angular vessel, most parcels shining clean blue while one in the line glows a faint wrong red, and a single scanning arch stands over the belt.](https://kinlane-images.s3.amazonaws.com/apievangelist/api-evangelist-images/newsletter/2026-10-12-skills-supply-chain.png)

The same week, the security industry started treating skills the way it treats packages.

- [VirusTotal published part three of its study of malicious agent skills](https://blog.virustotal.com/2026/10/from-automation-to-infection-part-iii.html), expanding from 3,016 skills to **35,878**, and naming, measuring and detecting the ones that carry a payload. [Tessl's talk from AI Native DevCon London makes the framing explicit: agent skills are supply chain components](https://tessl.io/blog/agent-skills-are-supply-chain-components). And [Port reviewed JFrog's Agent Skills Registry](https://www.port.io/blog/jfrog-agent-skills-registry) — once the artifact-repository vendors build a registry for something, it has become a dependency.
- The supply chain also got hit in the ordinary way. [Socket caught a compromised release of the TensorLake npm SDK](https://socket.dev/blog/tensorlake-compromise) — version 0.5.144, the SDK for an AI agent infrastructure company — carrying credential-stealing malware in a ChainDrop / Shai-Hulud attack. [Cloudsmith wrote it up too](https://cloudsmith.com/blog/tensorlake-sdk-npm-package-compromised).
- Meanwhile the legitimate skills kept shipping. [AWS published an agent skill for SageMaker-optimized inference](https://aws.amazon.com/blogs/machine-learning/new-agent-skill-amazon-sagemaker-optimized-generative-ai-inference-for-your-coding-agent/) through its Agent Toolkit. [Wagtail released agent skills for its developers](https://wagtail.org/blog/our-agent-skills-for-wagtail-developers/). [Sentry's plugin adds tools and skills to Unreal Engine's new Unreal MCP](https://blog.sentry.io/unreal-mcp-sentry/).

Put those together and the shape is clear. Skills are now distributed by servers, consumed by agents, and attacked like packages — all in the same week. The thing nobody has yet is the provenance layer between them: who wrote this skill, what it is allowed to do, and whether the copy your agent just loaded is the one that was published.

![A dark neon hall where a long row of identical angular doorways each carries the same glowing interlocking lock, and a small light-vessel passes through one of them as its lock opens in a clean ring of light.](https://kinlane-images.s3.amazonaws.com/apievangelist/api-evangelist-images/newsletter/2026-10-12-mcp-auth-patterns.png)

## MCP Authentication Grew Patterns

Two weeks ago the MCP auth story was a list of bugs. This week it was a list of patterns, which is what maturity looks like from the outside.

- [Okta showed how to build a secure TypeScript MCP client with Cross App Access](https://developer.okta.com/blog/2026/10/09/typescript-mcp-sdk-cross-app-access) — one enterprise app calling another on a user's behalf, with the identity provider in the middle rather than a token pasted into a config file.
- [Autodesk published authentication patterns for MCP servers on its platform](https://aps.autodesk.com/blog/authentication-patterns-aps-mcp-servers), following its earlier guide to building them. A platform writing down its auth patterns is a platform expecting other people to build servers on it.
- [MuleSoft wrote up a proxy pattern for governing remote MCP servers](https://blogs.mulesoft.com/dev-guides/governing-remote-mcp-servers-an-oauth-protected-resource-metadata-prm-proxy-pattern/) built around RFC 9728 protected-resource metadata — including the detail that the `resource` field must match the URL the client called byte for byte, which is exactly what breaks when you put a gateway in front.
- [WorkOS explained why Figma's MCP server returns 403s to legitimate clients](https://workos.com/blog/figma-mcp-403-forbidden-client-name): it admits clients by the `client_name` they send at dynamic registration, which proves nothing. And [AWS showed how to protect MCP endpoints at the edge with CloudFront and WAF](https://aws.amazon.com/blogs/networking-and-content-delivery/protect-mcp-endpoints-at-the-edge-with-amazon-cloudfront-and-aws-waf/).
- The reminder of why it matters: [UpGuard found exposed DBHub MCP servers leaking live databases to the open internet](https://www.upguard.com/blog/data-leakage-risks-with-dbhub-mcp-servers).

The protocol is also getting the long-running pieces it was missing. [Upstash explained MCP Tasks and the proposed MCP Events](https://upstash.com/blog/mcp-tasks-and-events) — durable long-running tools and agents that wake up — and [Google took the MCP Toolbox Java SDK to 1.0](https://cloud.google.com/blog/topics/developers-practitioners/announcing-mcp-toolbox-java-sdk-v10-agentic-data-access-for-the-enterprise/).

![A dark neon city road where two glowing gates stand side by side, one a single bright arch and the other a pair of interlocking rings, with small light-vessels passing through each.](https://kinlane-images.s3.amazonaws.com/apievangelist/api-evangelist-images/newsletter/2026-10-12-two-payment-rails.png)

***

**From the research: **[**Delivering a Modern API Integration Page with Arazzo**](https://papers.apievangelist.com/papers/arazzo-api-integration-page/?utm_source=apievangelist&utm_medium=email&utm_campaign=arazzo-api-integration-page&utm_content=midpoint)

You should not be building integrations anymore. You should be describing them. An integration is two ends and a wire — both ends already publish OpenAPI, and Arazzo is the wire. This paper turns the integrations page from a gallery of connectors into a directory of forkable workflows.

[Read the paper](https://papers.apievangelist.com/papers/arazzo-api-integration-page/?utm_source=apievangelist&utm_medium=email&utm_campaign=arazzo-api-integration-page&utm_content=midpoint)

***

## Agents Started Choosing Between Payment Rails

Last week the story was that you can now charge an agent per call. This week the people doing it started comparing notes.

- [Shveik built both x402 and MPP into one payment handler](https://blog.shveik.dev/x402-vs-mpp/) and wrote up how the traffic on their pay-per-call services actually split between the two. That is the first post I have seen measuring which rail agents choose when they are offered both.
- [MadeOnSol laid out MCP vs x402 vs API subscriptions for agents](https://madeonsol.com/blog/mcp-vs-x402-vs-api-subscription-ai-agent-data): MCP connects the agent to tools, x402 pays for an eligible call, a subscription authorizes sustained access. Those are three different contracts, and most providers are going to need to offer more than one.
- On the commerce side, [Constructor launched agentic checkout with Stripe](https://constructor.com/blog/introducing-agentic-checkout-from-constructor-and-stripe), and on-chain, a draft [ERC-8183 for agentic commerce](https://onchainagentintel.io/blog/erc-8183-agentic-commerce-erc-8004-reputation) proposes escrowed jobs with an evaluator attesting the result.

The common thread with the last section: every one of these rails needs to know which agent is paying. That is the problem I spent the week working on myself.

![A dark neon plain covered in thousands of tiny identical markers in rigid catalogue rows, with a handful of larger, distinct structures standing out among them.](https://kinlane-images.s3.amazonaws.com/apievangelist/api-evangelist-images/newsletter/2026-10-12-mcp-everywhere.png)

## MCP, Counted Honestly at a New Scale

The wider pull changed the MCP count more than anything else, and almost all of the change is one kind of source. Of **196** titles naming MCP this week, **113** come from three MCP directories the fix brought into view — vinkius.com, neuronto.com and agentalog.com — which publish one auto-generated listing per server, not posts. Four more are PyPI releases. Take those out and there are **79**, in a pool of 12,654 posts — about 0.6%. Last week was about 0.85% on a smaller, narrower pool. I am not going to call that a decline; the new coverage reaches a lot of blogs that were never going to write about MCP.

What the posts themselves said:

- The help-centre pattern is now everywhere — [Planhat published four pages on MCP plugins in a day](https://help.planhat.com/en/articles/17310389-planhat-os-mcp-plugins-overview), with QuotaPath, Nimble, Testlio and CompanyCam not far behind.
- The launches: [Bubble shipped Agent 2 and the Bubble MCP](https://bubble.io/blog/agent-2/), [PagerDuty the Rundeck MCP server](https://www.pagerduty.com/blog/announcements/the-rundeck-mcp-server-ai-where-your-automation-lives/), [Hashnode an MCP server for managing your blog](https://hashnode.com/blog/hashnode-mcp-server), and [Redis wrote up how it built its docs MCP for agents](https://redis.io/blog/how-we-built-the-redis-docs-mcp-for-agents/) so agents stop answering from pretraining.
- The "which surface" debate continued: [Playwright CLI vs MCP vs Kane CLI, compared on tokens, turns and reliability](https://www.testmuai.com/blog/playwright-cli-vs-mcp-vs-kane-cli/), [Semgrep on MCP giving agents tools while hooks give security control](https://semgrep.dev/blog/2026/mcp-vs-hooks-ai-agent-security), and [GC AI on what it learned building more than twenty agent connectors for legal work](https://gc.ai/blog/thinking-beyond-mcp). And from the people who generate SDKs for a living, [APIMatic asked what SDKs are for, twelve years on](https://www.apimatic.io/blog/what-sdks-are-for-12-years-on).

![A dark neon cartographer's table seen from above where a single luminous blueprint lies unrolled, a bright ring of nine connected nodes at its centre, and a small glowing badge resting at the edge of the map.](https://kinlane-images.s3.amazonaws.com/apievangelist/api-evangelist-images/newsletter/2026-10-12-from-my-desk.png)

## From My Desk: Know Your Agent, and the End of a Nine-Part Series

Across 12,767 third-party posts, OpenAPI appeared in two titles, both from Scalar, on generating it from [ASP.NET Core](https://scalar.com/blog/posts/2026-10-06-openapi-aspnet-core-dotnet-9-10) and from [FastAPI](https://scalar.com/blog/posts/2026-10-08-openapi-with-fastapi). The wider lens did not change that ratio at all.

- **Know Your Agent.** An agent cannot fill in a signup form, and the form was never really about authentication — it was how a provider found out who it was dealing with. So we shipped [the other half of Know Your Agent on APIs.io](https://apievangelist.com/2026/10/05/an-agent-that-signs-up-for-your-api-should-arrive-with-a-record/): an agent that signs up for your API should arrive with a record, every fact graded by how we know it, and two numbers — disclosure and verification — never collapsed into one. That is my answer to the question both sections above keep asking.
- **The Kin Score, facet by facet, finished.** Nine posts, one facet a business day: this week [access clarity](https://apievangelist.com/2026/10/05/the-kin-score-facet-by-facet-access-clarity/), [operational transparency](https://apievangelist.com/2026/10/06/the-kin-score-facet-by-facet-operational-transparency/), [create-or-update ergonomics](https://apievangelist.com/2026/10/07/the-kin-score-facet-by-facet-create-or-update-ergonomics/), [open source surface](https://apievangelist.com/2026/10/08/the-kin-score-facet-by-facet-open-source-surface/) and [regulatory posture](https://apievangelist.com/2026/10/09/the-kin-score-facet-by-facet-regulatory-posture/). Meanwhile [Nylas moved into first place on APIs.io at 93.2](https://apievangelist.com/2026/10/05/nylas-moves-into-pole-position-as-most-programmable-provider-on-apis-io/), past HubSpot, Salesforce and Harness, and I [ranked 79 industries by programmability](https://apievangelist.com/2026/10/09/seventy-nine-industries-ranked-by-programmability-and-the-opportunity-is-at-the-bottom/) — the opportunity is at the bottom of the table.
- **Specifications, from the inside.** [OpenAPI 3.2, a year on: most API tooling cannot read it](https://apievangelist.com/2026/10/09/a-year-after-openapi-3-2-most-api-tooling-cannot-read-it/), following [one in ten providers still publishing Swagger](https://apievangelist.com/2026/10/05/one-in-ten-api-providers-still-publish-swagger/). [Overlay 1.2 and the fourteen jobs we already use it for](https://apievangelist.com/2026/10/06/overlay-1-2-and-the-fourteen-jobs-we-already-use-it-for/), and [an overlay is a layer, not a diff](https://apievangelist.com/2026/10/08/an-overlay-is-a-layer-not-a-diff/). I sat in on [three specification calls in one Thursday](https://apievangelist.com/2026/10/08/three-specification-calls-in-one-thursday/), learned [what JOSE actually is](https://apievangelist.com/2026/10/08/what-i-learned-about-jose-on-todays-openapi-call/), and wrote about how [we used to ask the clients, and now we unleash the dogs](https://apievangelist.com/2026/10/08/we-used-to-ask-the-clients-now-we-unleash-the-dogs/).
- **And some opinions.** [It is called API design, you MCP hustlers](https://apievangelist.com/2026/10/05/it-is-called-api-design-you-mcp-hustlers/). [Twenty-four agent bundles and not one MCP server](https://apievangelist.com/2026/10/06/twenty-four-agent-bundles-and-not-one-mcp-server/). [984 companies hiring, and 99 of them say OpenAPI](https://apievangelist.com/2026/10/05/nine-hundred-eighty-four-companies-hiring-and-ninety-nine-of-them-say-openapi/). [Business capabilities provide the context you need for AI engineering](https://apievangelist.com/2026/10/08/business-capabilities-provide-the-context-you-need-for-ai-engineering/). [Who took my API catalog last week, and how much of it](https://apievangelist.com/2026/10/07/who-took-my-api-catalog-last-week-and-how-much-of-it/). [What I learned running the profiling pipeline on gpt-oss-120b](https://apievangelist.com/2026/10/07/what-i-learned-running-the-profiling-pipeline-on-gpt-oss-120b/).
- **The doors into APIs.io.** [Everything you can do with the APIs.io API](https://apievangelist.com/2026/10/05/everything-you-can-do-with-the-apis-io-api/) — 164 operations — and [the 141 tools in the APIs.io MCP server](https://apievangelist.com/2026/10/06/one-hundred-forty-one-tools-in-the-apis-io-mcp-server/). And a note on this newsletter itself: [Commune shipped an API, and our newsletters are moving onto it](https://apievangelist.com/2026/10/06/commune-shipped-an-api-and-our-newsletters-are-moving-onto-it/).

I want to end where the skills section left off. An agent loading a skill gets every signal that things are working — the skill loads, the tools appear, the calls succeed — and nothing tells it whether what it is running is what was published. The same goes for the agent paying for a call and the server deciding to let it in. Every layer of this stack is now trusting something it has not checked.

You only find that out by going and looking.

See you next week.

***

**🔌 **[**APIs.io — Know Your Agent**](https://apis.io/kya/?utm_source=apievangelist&utm_medium=email&utm_campaign=apisio-kya&utm_content=footer) — If you run an API and agents are starting to sign up for it, this is the part of APIs.io I would point you at this week. An agent publishes an A2A card, registers it, and builds a record with us — who operates it, how to reach them, what it is for — with every fact graded verified, attested, declared or absent. When that agent signs up for your API, the record can travel with it, so you get back the thing the signup form used to give you: who you are dealing with.

[Start at apis.io/kya](https://apis.io/kya/?utm_source=apievangelist&utm_medium=email&utm_campaign=apisio-kya&utm_content=footer-link) to see what a record holds.

***

## Discussion

No replies yet.