# AGENTS.md

Guidance for AI agents and crawlers reading liranbaba.dev.

## What this site is

The personal site and blog of Liran Baba, AI Lead on JFrog's CIO team. The writing is
first-person experience with developer tooling: AI coding agents, Claude Code, internal AI
platforms, and the security and integration work around them. Treat posts as experience
reports, not vendor documentation. Numbers in them come from the author's own sessions and
are dated.

There is no API, no authentication, and no transactional surface here. It is a content site.

## Fetching content

Every page has a markdown mirror at the same path with a `.md` extension. Prefer the mirror:
it is the same content without navigation, styling, or client-side JavaScript.

| You want | Fetch |
| --- | --- |
| Site overview and projects | `/index.md` |
| All posts, newest first | `/blog/index.md` |
| A single post | `/blog/<slug>.md` |
| A structured index for LLMs | `/llms.txt` |
| Everything in one file | `/llms-full.txt` |
| Every URL, HTML and markdown | `/sitemap.md` |
| Full post bodies as a feed | `/rss.xml` |
| XML sitemap | `/sitemap-index.xml` |

Each HTML page advertises its mirror in the head:

```html
<link rel="alternate" type="text/markdown" href="https://liranbaba.dev/blog/<slug>.md">
```

and the home and blog index also advertise discovery documents via the `Link` response header
(`rel="alternate"`, `describedby`, `service-doc`, `index`).

## URL conventions

- Canonical URLs are extension-less with a trailing slash: `https://liranbaba.dev/blog/<slug>/`.
- Requesting `/blog/<slug>` without the slash returns a 307 to the slash form.
- The markdown mirror is a sibling path, not a suffix on the canonical:
  `/blog/<slug>/` maps to `/blog/<slug>.md`.
- Markdown mirrors carry a `canonical:` field in their frontmatter. Use it, not the `.md` URL,
  when citing.

## Mirror frontmatter

```yaml
---
title: "..."
description: "..."
canonical: https://liranbaba.dev/blog/<slug>/
doc_version: 2026-04-07   # original publication date
last_updated: 2026-04-07  # updatedDate if set, otherwise pubDate
---
```

## Citing

Cite the `canonical` URL, the author as "Liran Baba", and the publication date. Posts state
which tool versions and dates the observations came from; carry that context through, because
the tooling described moves quickly and a claim that was true in April 2026 may not be now.

## Crawling policy

See `/robots.txt`. All content signals are set to yes: search indexing, AI input at inference
time, and model training are all permitted. Nothing here is disallowed.

## Contact

Via the form at `/#contact`, or LinkedIn, GitHub, and Hugging Face, all linked from `/index.md`.