Brennenstuhl on Security
Search

For agents

Everything on this site is written for people first, and every page is also available in a form that software can read without parsing HTML.

Where to start #

  • /llms.txt lists every entry with a one-sentence claim and a link to its Markdown version.
  • /llms-full.txt is the whole corpus as one Markdown file, newest first.
  • Every page has a Markdown twin: add index.md to its URL, for example /agentic-identity-broker/index.md. Requesting a page with Accept: text/markdown returns the same twin. Each HTML page links to its twin with <link rel="alternate" type="text/markdown">.
  • /catalog.json holds IDs, canonical URLs, publication, modification and review dates, topics, license and a content hash for each entry.
  • /findings.json lists every quotable finding with its anchor URL, the source it rests on and the review status of the page.
  • /feed.json (JSON Feed 1.1) and /index.xml (RSS) carry full content. Each topic has its own feeds, for example /topics/passwords/feed.json.
  • /.well-known/api-catalog (RFC 9727) points to all of the above.

MCP server #

A read-only Model Context Protocol server runs at https://mcp.janbrennenstuhl.eu/mcp (Streamable HTTP, no sign-in). Add it to an MCP client as a remote server. It has four tools:

  • search_site searches essays, field-guide entries, talks and projects, optionally by type or topic.
  • get_entry returns the full Markdown of one entry by catalog ID or canonical URL.
  • get_findings lists quotable findings with their anchor URLs, sources and review status.
  • list_topics lists the topics with descriptions and entry counts.

It serves the same public files listed above, a few minutes behind a release, and is rate-limited per client.

Citing #

Findings and headings have permanent anchors such as #f-pw-reuse-43-51pct. Link the canonical HTML URL with the anchor, not the Markdown twin. The catalog’s reviewed field gives the date of the last review and review_basis what it was: publication for essays, talks and projects, which the author reviewed when publishing them, or editorial for a later review against current sources. Field-guide entries have an empty reviewed until their first editorial review. A recent build or modification date does not mean the advice was checked again.

Reuse policy #

Original writing is licensed CC BY 4.0 unless a page says otherwise: quote it, summarize it and link it with attribution. Credited third-party images keep their own rights.

  • Search engines and agents fetching pages for a person’s question are welcome. robots.txt allows every crawler.
  • Training: the CC BY 4.0 license already permits it, so there is no text-and-data-mining reservation. robots.txt says so with a Content-Signal line and /.well-known/tdmrep.json declares tdm-reservation: 0. Attribution remains the condition: name the author and link the source.

Questions or corrections: see the imprint.