Claude Code v2.1.152 ships /code-review --fix, disallowed-tools frontmatter, /reload-skills
The Claude Code v2.1.152 release is a developer-experience update with a sharper edge than usual. The headline change is /code-review --fix: when invoked, Claude Code now applies review findings directly to the working tree rather than printing them as suggestions. /simplify is routed through the same path, meaning two of the most-used review-style commands now produce diffs by default. For agentic-coding power users this is a meaningful workflow change — review and refactor become single-shot operations instead of review-then-prompt sequences.
The security-relevant change is disallowed-tools frontmatter. Skills and slash commands can now declare, in their YAML frontmatter, specific tools that the command is forbidden from invoking. That gives skill authors a way to restrict, say, a documentation-generation skill from ever invoking shell execution, even if the underlying Claude model would otherwise be willing to use it. In the context of Anthropic's broader Mythos-class capability work and the enterprise-security integration push (CrowdStrike, Wiz et al), the disallowed-tools mechanism is the developer-surface analog of those enterprise controls.
The small but quality-of-life-significant change is /reload-skills, which rescans skill directories without restarting Claude Code. Source C's coverage of an agent profiler ('AgentLens') called out that authors are loading 'hundreds of skills' into long-running sessions; until now, iterating on those skills meant a full restart of the session and the loss of context. /reload-skills closes that gap.
These changes land in a week when Boris Cherny, Claude Code's lead, was on X pitching 'auto mode' — running Claude Code without per-action permission prompts — as his #1 productivity tip. The combination is consistent: Anthropic is pushing Claude Code toward more autonomous defaults while giving skill authors and enterprise admins finer-grained control over what those defaults can actually do. The Mythos rollout question hangs over all of this — disallowed-tools is exactly the kind of guardrail Anthropic will need to point at when explaining how a more capable model can ship safely inside Claude Code.