fix(agent): stop sending every page view to a third-party avatar CDN (#1226)

* fix(agent): stop sending every page view to a third-party avatar CDN Eight avatar SVGs were loaded from api.dicebear.com on every render. In an accounting product that meant every authenticated page view told a third party who was looking at it, from a domain we do not control, on the path of a logged-in surface. A firewalled or self-hosted install showed no faces at all. The SVGs are now generated once and served from public/agent-avatars. Each entry records the seed it came from, so the set can be regenerated reproducibly, and the command to do it is in the file. The licence question that made this look like a founder decision resolved itself on inspection: Notionists is by Zoish under CC0 1.0, public domain, no attribution required. Confirmed on dicebear.com/licenses and, more usefully, in each downloaded file's own RDF metadata, so the terms travel with the asset rather than living in a commit message. Tests pin the properties that matter rather than the file list: no entry may be a remote URL, every entry must have a file behind it, and no shipped SVG may carry an <image href>, a url(https://…), an xlink:href or a <script>, since self-hosting a file that then phones home would reintroduce exactly the request this removes. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(agent): assert the property, not a list of elements, for avatar externals The external-reference check enumerated <image href>, url(https://…) and xlink:href, which left <use href>, <feImage href> and scheme-relative //host through: exactly the requests the guard claims to prevent, via elements it happened not to list. That is how this sort of allowlist rots. It now strips the parts that legitimately carry URLs and are never fetched (the RDF metadata block, xmlns declarations) and then asserts that NOTHING in what remains points off-origin. Verified by injecting each of the four bypasses into a real asset and confirming the test fails on all of them. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
2026-07-27 15:35:14 +02:00
parent 248d98bd7e
commit 2da96c0be2
12 changed files with 131 additions and 30 deletions
@@ -599,3 +599,4 @@ One line per decision: `[YYYY-MM-DD] <decision>: <why>`. Appended by agents and
 [2026-07-27] Analytics is hosted-only and explicitly so: isAnalyticsEnabled() short-circuits on NEXT_PUBLIC_SELF_HOSTED and no __NEXT_PUBLIC_POSTHOG_*__ sentinel was added to Dockerfile/docker-entrypoint.sh. Recapt reached the same outcome only by accident (its env var was simply missing from the sentinel list), which meant a self-hosted operator could never have configured it and nobody had decided that on purpose. An AGPL operator's users should not be reported to our project.
 [2026-07-27] Removing Recapt made lib/support/submit-feedback.ts single-channel: a failing /api/support/contact now surfaces as a real error instead of being masked by Recapt reporting success on its own channel. That is the correct behaviour (silently "succeeding" while the message reached nobody was worse) and the Resend path is solid. A non-blocking posthog.capture('support_feedback_submitted') replaces the useful half of the Recapt channel by putting the submission on the user's timeline next to the session replay; it deliberately carries no message body, since free text is user content and would be PII in an event property.
 [2026-07-27] vitest.config.ts now aliases 'server-only' to tests/stubs/server-only.ts. Its real entry point throws unconditionally (Next.js swaps it out at bundle time; Vitest cannot), so the moment a server-only module entered the test graph it broke 48 test files at import. app/(dashboard)/request-context.ts was already carrying the same latent trap and had simply never been imported by a test.
+[2026-07-27] Self-hosted the dicebear Notionists avatars under public/agent-avatars instead of loading api.dicebear.com per render: the licence turned out to be CC0 1.0 (verified on dicebear.com/licenses AND in each file's own RDF metadata), so there was no licence decision to escalate, and the CDN was sending every authenticated page view's IP and referer to a third party while breaking firewalled/self-hosted installs entirely.