Simon Willison: Practical LLMs & AI Engineering

Ravenza Editorial

41 installs · 80 sources · Updated 8/18/2026

Simon Willison is a developer and researcher who documents hands-on experiments with language models as they evolve. This Skill makes his work searchable across prompting, tool use, agents, coding workflows, model behavior, open-source utilities, and prompt-injection security.

simon-willisonllm-engineeringcoding-agentsprompt-injectionmcplocal-modelsstructured-outputretrievalevaluationtool-use

Example questions

→Design a reproducible evaluation for two coding models.

→Why does this local model overthink simple tasks?

→What should an LLM experiment log?

→Design a bounded tool loop for this agent.

→Threat-model an MCP server for prompt injection.

→Does this agent create the lethal trifecta?

Related Skills

Sources

80 web pages

Qwen 3.8 27B is excellent, but it defaults to wildly overthinking thingssimonwillison.net One-shotting a Raccoon Heist game using Claude Fable 5simonwillison.net New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter loggingsimonwillison.net Stateless MCP has recaptured my interestsimonwillison.net OpenAI's accidental cyberattack against Hugging Face is science fiction that happenedsimonwillison.net A Fireside Chat with Cat and Thariq from the Claude Code teamsimonwillison.net

Update history

Skill details refreshed · 8/18/2026

The title, description, or cover was improved.

Initial release · 8/17/2026

First published on Ravenza.