reviewlisted
Install: claude install-skill x-mesh/xm
<Purpose>
x-review takes a PR diff, file, or directory as input and runs multiple review agents in parallel. Each agent reports findings from a dedicated perspective (security, logic, performance, test coverage) in severity + file:line format. The leader then produces a consolidated report with an LGTM / Request Changes / Block verdict.
</Purpose>
<Use_When>
- User wants to review a PR, file, or directory
- User says "review", "code review", "check PR", "analyze diff", "review"
- User says "check security vulnerabilities", "find performance issues", "check test coverage"
- Other xm skills need a code quality gate
</Use_When>
<Do_Not_Use_When>
- Simple single-line questions that don't need multi-agent review
- Structured problem solving (use x-solver instead)
- Full project lifecycle management (use x-build instead)
</Do_Not_Use_When>
# x-review — Multi-Perspective Code Review
Parallel review orchestrator built on Claude Code native Agent tool.
No external dependencies. Only requires `git` and `gh` CLI.
## Execution Boundary — Headless / Delegated Runtimes
Before Phase 1, inspect the tools and execution context actually available in this invocation.
- If native Agent/collaboration fan-out is unavailable, or the prompt says this is a headless,
rescue, delegated, leaf, or single-agent run, **do not spawn agents and do not call any
collaboration wait primitive**. Run a single-pass review in the current process: inspect the
complete frozen target once with the `correc