d415294be021ec2dbc3d9bfacc43371b7e9c0080
docs 06 §7 has specified `public/.well-known/security.txt` since the start and it was never created — there is currently no machine-readable way to report a vulnerability, and no SECURITY.md for GitHub's Security tab either. - `public/.well-known/security.txt` — RFC 9116 valid (Contact + Expires are the required fields; Expires is 10.7 months out, inside the 12-month guidance). Contents follow docs 06 §7 verbatim, plus a `Policy:` line. - `SECURITY.md` at the repo root — reporting address, response expectation, and an explicit in/out-of-scope list so scanner-output reports can be declined without a discussion. Includes what the site actually handles (no accounts, no trackers, one Turnstile-gated form) since that governs real impact. - `public/_headers` — `text/plain; charset=utf-8` for security.txt. Wrangler already infers this from the extension; the entry is belt-and-braces. - `public/robots.txt` — the AI-crawler block listed 5 user-agents and had gone stale. Now 25, covering OpenAI (incl. OAI-SearchBot / ChatGPT-User), Anthropic, Perplexity, Apple, ByteDance, Amazon, Meta, Cohere, Mistral, Diffbot, AI2 and friends. Content is All Rights Reserved, so this mirrors the Cloudflare "Block AI Scrapers" toggle. Verified `.well-known/` survives the pipeline: Astro copies it to dist/client/.well-known/security.txt, the generated `_headers` carries the entry, and `.assetsignore` only excludes wrangler.json and .dev.vars — so wrangler will upload it. Confirmed with curl after deploy. Note: the `Policy:` URL resolves once the repo is public. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Languages
TypeScript
51.9%
Astro
19.4%
Svelte
16.3%
JavaScript
9.9%
CSS
2.5%