ARCH-4: Document LLM/code-generation threat boundaries #4

Closed
opened 2026-05-06 21:14:39 -07:00 by jwilger · 0 comments
Owner

Roadmap key: ARCH-4

Depends on: ARCH-1

Acceptance criteria

  • Identifies untrusted LLM output, prompt/vector memory sensitivity, event payload PII, generated code artifacts, hot-load, external API calls, command authorization, and validation/test/compile gates.
  • Documents Phoenix baseline trust boundaries introduced by BASE-1, including signed cookie sessions, CSRF, LiveView socket session handling, dev-only dashboard/mailbox routes, runtime secret loading, and local Postgres access.
  • Records production-hardening decisions or follow-up work for secure session cookies, host/header assumptions around SSL redirects and health-check exclusions, and Postgres TLS requirements for remote production databases.

Source

Created from .kilo/plans/1778116681638-nimble-star.md.

Roadmap key: `ARCH-4` Depends on: [ARCH-1](https://git.johnwilger.com/jwilger/anything_app/issues/1) ## Acceptance criteria - Identifies untrusted LLM output, prompt/vector memory sensitivity, event payload PII, generated code artifacts, hot-load, external API calls, command authorization, and validation/test/compile gates. - Documents Phoenix baseline trust boundaries introduced by BASE-1, including signed cookie sessions, CSRF, LiveView socket session handling, dev-only dashboard/mailbox routes, runtime secret loading, and local Postgres access. - Records production-hardening decisions or follow-up work for secure session cookies, host/header assumptions around SSL redirects and health-check exclusions, and Postgres TLS requirements for remote production databases. ## Source Created from `.kilo/plans/1778116681638-nimble-star.md`.
Sign in to join this conversation.
No description provided.