# Changelog

## Score

- CAI 38 → 38 (+0.1)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.2) — scores are not directly comparable.

## Lenses

- Code Health 56 → 56 (+0.0)
- Architecture 100 → 100 (+0.0)
- Maturity 61 → 61 (+0.0)
- Readiness 41 → 41 (+0.0)
- Security 74 → 75 (+1.4)
- Accessibility 20 → 20 (+0.0)
- Performance 85 → 85 (+0.0)

## Resolved (6)

- Documentation: no architecture or design documentation
- Inconsistency in `global_subscribe` signature: `Redis.global_subscribe` accepts a block (`blk`), while `Memory` and `Postgres` do not show a block parameter in their signatures. This suggests a lack of uniform interface across backend implementations for the same operation.
- Low CVE: [GHSA redacted] (pnpm-lock.yaml)
- Multiple distinct types implement encoding/decoding logic with inconsistent naming and signatures. `MarshalCodec`, `PackedString`, `StringHack`, and `MessageBus.Codec.*` all define `encode` and `decode` methods, but they operate on different data structures (hashes vs payloads) and are not unified under a single interface or base class in the public API surface. Furthermore, `MessageBus.Message` has its own `encode`/`decode` pair, and a separate `encode_without_ids` method, creating confusion about which encoder to use for which domain.
- Off-boarding risk: anonymized user #1
- The backend implementations (`Memory`, `Postgres`, `Redis`) duplicate the interface defined in `Base`, but `Base` also exposes `last_id` and `last_ids`. While this is technically inheritance, the public API surface exposes both the base and concrete implementations with identical signatures. If `Base` is abstract, it should not be part of the public API surface alongside its implementations, or the implementations should not expose these methods if they are strictly internal to the bus implementation.

## New (2)

- High CVE: [GHSA redacted] (pnpm-lock.yaml)
- Off-boarding risk: anonymized user #1

## Architecture

- Unchanged — 0 containers · 1 contexts · 0 edges
