Add Continue AI checks for automated PR review

[sestinj]

Summary

Adds Continue AI checks to run on every PR, providing automated code review for areas that require judgment beyond what linters and type checkers can catch.

Checks

• Sync Correctness — Verifies sync logic preserves the append-only, idempotent, convergence-over-coordination guarantees from docs/SYNC.md. Catches source-of-truth violations, idempotency breaks, and cursor/watermark issues.
• Error Handling Patterns — Ensures consistent use of anyhow vs thiserror, proper error context propagation, no silent failures, and daemon resilience (one session's error shouldn't crash the loop).
• Security & Auth Review — Checks token handling, credential storage encryption, secret leakage in logs, HTTPS enforcement, and auth flow correctness.

Why these checks?

These target the three highest-impact areas identified from codebase analysis:

1. Sync is the critical path — bugs here cause silent data loss in a background daemon
2. Error handling consistency — the daemon can't show errors to users, so proper logging and resilience are essential
3. Security surface area — OAuth tokens, encrypted credential storage, and cloud transmission require ongoing vigilance

Learn more: docs.continue.dev

---

Continue Tasks: ✅ 2 no changes — View all