For AI Agents
Skills, tools and structured data for LLMs and agents working with this site. Nothing here needs an account or an API key.
Agents: start at agents.danielrosehill.com.
It is the machine-readable version of this site, generated at each build: llms.txt, JSON indexes
of my projects, models, datasets and plugins, the skills below, and the MCP server.
Agent Skills
Three skills in the Agent Skills
format: a folder with a SKILL.md holding YAML frontmatter (name,
description) and instructions. Load one into Claude, or any agent that reads
the format, and it knows how to do that task here.
contact-daniel Send Daniel a message on a user’s behalf through the contact form endpoint, or hand the user email or a prefilled form.
book-meeting-with-daniel Request a video or phone call: the user proposes times and Daniel confirms by email. Includes a prefilled booking link.
about-daniel Who Daniel is and where to find his repositories, Hugging Face models and datasets, system prompts, podcast and CV.
Machine-readable index, per the Agent Skills discovery draft (v0.2.0), with a sha256 digest
for each file: /.well-known/agent-skills/index.json.
To install a skill by hand, save its SKILL.md as
~/.claude/skills/<name>/SKILL.md.
Actions you can take without an account
Both actions send email to a real person. Confirm the exact content with your user and use their real name and email. Send once.
Contact Daniel
- Programmatic:
POST https://www.danielrosehill.com/api/contactas form data withname,email,subject,messageand the_tstimestamp. The site’s own form uses this endpoint. Full field list and spam rules: contact-daniel. - Prefilled form for the user:
/contact/form?name=…&email=…&subject=…&message=… - Email: public@danielrosehill.com
Request a meeting
- Meetings are requested, not instantly booked. The user proposes times, and Daniel replies by email with a calendar invitation.
- Programmatic: the same endpoint with
formType=Meeting request,preferredTimes,timezoneandtopic. See book-meeting-with-daniel. - Prefilled link for the user:
/book-meeting?name=…&email=…&duration=30%20minutes&preferredTimes=…&topic=…
WebMCP (agents inside the browser)
Every page on this site registers read-only
WebMCP
tools on document.modelContext (navigator.modelContext in early
Chrome previews). A browser agent that supports WebMCP sees them as soon as a page loads,
and can call them instead of scraping the page. None of them submit a form or send
anything off the site.
get_site_overview | The llms.txt overview. Call this first. |
search_projects | Keyword search over about 1,400 GitHub repositories and Hugging Face items. |
list_project_categories | Project categories with counts. |
get_ai_focus | AI focus areas, flagship models and datasets, Hugging Face collections. |
list_ai_models_and_datasets | Every Hugging Face model, dataset and space. |
get_podcast_info | My Weird Prompts: episode count and audience figures. |
search_system_prompts | Search about 1,300 public system prompts. |
list_websites | Daniel’s other websites and blogs. |
get_contact_options | Email, contact form and its prefill parameters, profiles. |
get_booking_options | How to request a meeting, and a prefilled /book-meeting link. |
list_agent_skills | The Agent Skills on this page, optionally with their full text. |
read_page | Read any page on this site as text. |
In Chrome, WebMCP needs the “WebMCP for testing” flag or the Model Context Tool Inspector extension.
Remote MCP server
A public Model Context Protocol server (streamable HTTP, no authentication) for MCP clients such as Claude:
https://mcp.danielrosehill.com/mcp
Tools: about_daniel, ask_about_daniel, get_resume, download_cv, search_projects, get_blog_posts, get_recommendations, send_message, request_meeting. send_message and request_meeting post to the same endpoint as the contact and meeting skills above.
llms.txt and structured data
Static files, rebuilt on every deploy. Prefer the JSON to scraping HTML.
/llms.txtConcise site overview, identity and section map./llms-full.txtExtended bio, expertise and project taxonomy./ai.txtAI permissions: crawling, indexing and training are allowed./webmcp/projects.jsonEvery public GitHub repository and Hugging Face item: title, description, subtype, categories, tags, URLs./webmcp/ai.jsonAI focus areas, flagship work, model and dataset cards, all Hugging Face items, podcast figures./system-prompts/index.jsonAbout 1,300 system prompts with full text (about 2.2 MB)./data/websites.jsonOther websites and blogs./daniel-rosehill.vcfContact card (vCard)./sitemap-index.xmlXML sitemap./rss.xmlRSS feed.
Quick facts
Daniel Rosehill is an AI developer and technical communicator based in Jerusalem, Israel. He works on multi-agent orchestration, Claude Code workflows, MCP servers, speech recognition for English–Hebrew code-switching, and the documentation that makes those systems usable. Active across GitHub, npm and Hugging Face. Available for consulting through DSR Holdings and its AI consultancy Carrot Cake AI.
Base URL: https://www.danielrosehill.com. The apex danielrosehill.com redirects there.
Content licence CC BY-SA 4.0.