Skip to main content
Not really bugs. Dead code, pointless types, a test-coverage confession. If you’re reading the codebase and something looks off, it might be here.

​
Dead variable in the planner

Slug: tasks-recorded-is-dead-state There’s a tasksRecorded: string[] in the planner that gets populated inside buildSkillTool — and then never returned, never persisted, never read by anything. Pure dead code. What to do. Nothing. Delete it on the next cleanup pass.

​
A conditional type that isn’t

Slug: map-finish-reason-pointless-ternary mapFinishReason in gateway/src/session/prompt.ts has a parameter type of StreamEvent["type"] extends "finish" ? any : any. Which — no matter how you squint at it — is any. The conditional adds zero information. Someone was probably partway through refactoring. What to do. Simplify on the next cleanup pass.

​
Effect.promise eats rejections silently

Slug: db-effect-promise-swallows-errors Two spots in the DB layer use Effect.promise(...). Subtle thing about Effect.promise: when the underlying promise rejects, it treats that as a defect, not an Effect error. Which means transient Supabase failures at those call sites can silently resolve without the caller ever seeing them. What to do. Audit every Effect.promise call. Prefer Effect.tryPromise with an explicit catch for anything non-trivial.

​
The honest test-coverage backlog

Slug: test-coverage-gaps Less a bug, more a confession. The test suite doesn’t cover, and probably should:
  • Concurrent /plan requests end to end (only the pubsub filter is tested).
  • Compaction correctness on long multi-pass sessions.
  • Revert.
  • SSE frame ordering under load.
  • Non-English payloads.
  • Sessions larger than the 1000-row pagination limit.
  • Missing bearer_env env vars.
  • Aborted requests propagating to the Bindu client.
  • The snake_case flip on tasks/cancel.
What to do. Nothing — this is the backlog. If you’re fixing one of the items on the other severity pages, add a test for it in the same PR. That’s how this list shrinks.