ClawVault v0.3.0
This release adds end-to-end Claude Code guard integration and improves OpenClaw runtime protection, making
ClawVault able to monitor and control both user prompts and tool/runtime actions with the same detector,
RuleEngine, guard modes, audit store, and dashboard semantics.
Highlights
- Added Claude Code
UserPromptSubmithook support to scan prompts before they are sent to the model. - Added Claude Code
PreToolUseruntime-action guard support for tool execution checks. - Added safe Claude Code hook installer and setup script with dry-run, install, status, and uninstall flows.
- Improved OpenClaw Runtime Action Guard integration and adapter coverage.
- Updated Dashboard Audit Log and Protection Monitor to show persisted prompt/tool guard events.
- Added OpenAI-compatible sanitize adapter support for OpenClaw deployments.
- Expanded tests for Claude Code hooks, runtime action APIs, dashboard audit visibility, and OpenClaw guard
behavior. - Updated English and Chinese README documentation for the new integration flow.
Notes
Claude Code hooks fail closed when the local ClawVault guard API is unavailable. Keep ClawVault running while
hooks are installed.
Install preview:
./scripts/setup-claude-code.sh --dry-run
Install after review:
./scripts/setup-claude-code.sh --yes