Skip to content
View kristian-presso's full-sized avatar
1 follower · 2 following

Achievements

Achievement: Pull Sharkx4Achievement: Pair Extraordinairex4Achievement: QuickdrawAchievement: YOLO

Achievements

Achievement: Pull Sharkx4Achievement: Pair Extraordinairex4Achievement: QuickdrawAchievement: YOLO

Block or report kristian-presso

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
kristian-presso/README.md

Hi, I'm Kristian 👋

I build AI agents with Claude that run real businesses, then open-source the parts that work.

Ten years in enterprise B2B taught me that a clever system is worth nothing until it survives contact with real customers. So everything here is tested: every skill I publish is scored against plain Claude, and only ships if it makes Claude better.

What I build

  • 🧰 claude-skills - 21 Claude skills in five plugins: MCP servers and agent security, outbound sales, AI search visibility, running a UK company, and the operator discipline that stops an agent doing two hours of good work on the wrong brief. Each one beats plain Claude in evals.
  • 🛫 connector-preflight - check an MCP server against Claude's connector directory rules before you submit: the OAuth discovery chain, tool annotations and the review criteria, each failure with its fix and its source. CLI, GitHub Action and Claude Code plugin.
  • 🤖 MCP servers in production - for event ticketing, home search and an AI co-founder, with real users, real payments, OAuth and approval gates on anything public, costly or irreversible.
  • ✍️ Guides - how it is built: tool design, agent memory, approval gates, auth for agents, multi-agent orchestration with Claude Code.

How I work

  • Agents get clear roles, skills hold the judgement, and memory carries what they learn.
  • A human approves anything public, anything involving money and anything that cannot be undone.
  • If it is not measured, it is not done. Evals before launch, tests on every push.

Before AI

Enterprise customer success at IDC, Gartner and Amplitude, looking after Fortune 500 accounts including UBS, LEGO, Danske Bank and Skoda. 118%+ net revenue retention and £1.5m in expansion revenue. I know what it takes to get a new technology adopted inside a large company, not just demoed there.

Want to work together?

I'm interested in teams putting Claude and MCP to work inside real companies: designing the agents, getting them through security review, and proving they paid off. If that is what you are building, I would like to hear from you.

LinkedIn · kristianpapadakis.com · London, UK

Pinned Loading

  1. claude-skills claude-skills Public

    Claude skills and plugins from the agents that run my business. Every skill tested against plain Claude. Two commands to install. Free, MIT.

  2. PressoNetwork/connector-preflight PressoNetwork/connector-preflight Public

    Check a remote MCP server against the Claude connectors directory requirements before you submit. Read-only, zero dependencies. CLI, GitHub Action and Claude Code plugin.

    JavaScript