# Changelog

## Baseline

- First survey — no prior run to compare against. CAI 41.0.

## Lenses

- Code Health 39.7
- Architecture 94.9
- Maturity 82.3
- Readiness 50.0
- Security 77.9
- Accessibility 24.7
- Performance 97.1

## Changes since last survey

- 300 commits — 196 feature/other, 104 fixes

## By area

- (repo) — 104 commits
- cmd/rampart — 55 commits
- (root) — 44 commits
- internal/engine — 19 commits
- .github/workflows — 10 commits
- internal/audit — 10 commits
- internal/plugin — 10 commits
- internal/securefile — 6 commits
- docs/assets — 5 commits
- assurance/integrations.yaml — 3 commits
- docs-site/getting-started — 3 commits
- docs-site/integrations — 3 commits
- sdks/python — 3 commits
- .github/ISSUE_TEMPLATE — 2 commits
- .github/dependabot.yml — 2 commits
- docs-site/features — 2 commits
- docs-site/index.md — 2 commits
- docs/index.html — 2 commits
- internal/bridge — 2 commits
- internal/notify — 2 commits

## Notable commits

- fix: Merge main into staging after the v1.9.6 hook fixes
- fix: Merge pull request #474 from peg/codex/revert-467
- fix: Merge pull request #489 from peg/codex/1.10-release-fixes
- fix: Revert "fix(setup): fail closed when the Claude Code hook cannot run" (#467)
- fix: fix(approval): bind redacted review to immutable actions
- fix: fix(approval): capture input once and preserve run scope identity
- fix: fix(audit): cancel witness waits on shared writer locks
- fix: fix(audit): classify records by event id in the reader, as the verifier does
- fix: fix(audit): close file handles even when final sync fails
- fix: fix(audit): fail verification when a record holds content the hash skips
- fix: fix(audit): report malformed records instead of skipping them silently
- fix: fix(audit): roll back incomplete appends after write failures
- fix: fix(audit): skip chain-continuation headers when reading events
- fix: fix(claude-code): keep Rampart's hook decisive and doctor visible
- fix: fix(claude-code): plan fail-closed for a new PreToolUse hook even with leftover Post hooks
- fix: fix(cli): guide first-run users toward protect
- fix: fix(cli): keep counting audit events beside a malformed record
- fix: fix(cli): label ask decisions in rampart test
- fix: fix(cli): redact final command error output
- fix: fix(cli): replace inherited diagnostic backups on Windows
- …and 280 more

## Architecture

- 0 containers · 1 bounded contexts · 0 dependency edges (baseline)
