Static site (plain HTML/CSS/JS, no build step) published via GitHub Pages at https://virtual-engineer.dev/
index.html— landing pagedocumentation.html— documentation pageintegrations.html— integrations pagesecurity.html— security pageabout.html— about page404.html— not-found page, served by GitHub Pages for unknown pathsrobots.txtandsitemap.xml— crawler discoveryfavicon.ico— site faviconassets/— stylesheets, scripts, and images shared across pages
These files make the site legible to AI agents and answer machine-readable discovery checks (see https://agentgrade.com/):
llms.txt— content index and operating manual for agentsllms-full.txt— full concatenated text of the site, for non-browsing LLMsagents.txt— agent access policy (what agents may and may not do here)feed.xml— RSS feed of site updates.well-known/agent-card.json— A2A agent cardscripts/build-llms-full.py— regeneratesllms-full.txt(see below)
No dependencies or build step are required. Serve the folder with any static HTTP server from the repo root, for example:
python3 -m http.server 8000Then open http://localhost:8000 in a browser.
Opening the HTML files directly via file:// also works for most changes,
but a local server is recommended since some browsers restrict local script
loading over file://.
Pushes to main are published automatically to GitHub Pages.
The custom domain should have GitHub Pages' Enforce HTTPS setting enabled,
which also makes http:// requests redirect to https://.
llms-full.txt is generated, not hand-written. It concatenates the readable
text of every page, plus the documentation content from assets/docs-data.js.
After changing page content or documentation data, run:
python3 scripts/build-llms-full.pyThe script needs Python 3.8+ and no third-party packages. Commit the regenerated
llms-full.txt along with the content change.
There is no automated test suite. For a visual smoke check, load every page over
http://localhost:8000, exercise navigation and the theme toggle, and check
desktop and mobile layouts. For the agent-readiness checks, scan the published
site:
curl "https://agentgrade.com/api/scan?url=https://virtual-engineer.dev"