nord-reviewlisted

# nord-review — multi-agent adversarial review Invoking this skill IS opt-in to multi-agent orchestration. Run the Workflow tool with the script below. First, inline, establish the review target (default: `git diff HEAD`; or the files/PR the user named) and pass it as `args.target`. Why this over a single reviewer: one model misses failure modes. Independent per-dimension reviewers catch more; adversarial verification kills plausible-but-wrong findings before they reach the user. After the workflow returns, present `findings` grouped by severity (critical→low) as `path:line — title. fix.`; if `--fix` was requested, apply the confirmed findings. ```javascript export const meta = { name: 'nord-review', description: 'Multi-dimension adversarial code review', phases: [ { title: 'Review', detail: 'parallel per-dimension reviewers' }, { title: 'Verify', detail: 'adversarially verify each finding' }, ], } const target = (args && args.target) || 'the current git diff (run: git diff HEAD)' const DIMENSIONS = [ { key: 'correctness', prompt: 'Logic bugs, off-by-one, null/undefined, error handling, race conditions, broken edge cases.' }, { key: 'security', prompt: 'Injection, authz/authn gaps, secret leakage, unsafe deserialization, path traversal, SSRF.' }, { key: 'performance', prompt: 'N+1, needless allocation, sync-in-hot-path, missing indexes, accidental O(n^2).' }, { key: 'reuse', prompt: 'Duplicated logic, reinvented stdlib/lib, needless abstra