Agents

Specialized subagents for PR review and UX audit tasks. Review skills such as /trackbed:pr:code-review and /trackbed:pr:product-review invoke them, and you can also call them directly through the Task tool. Code review records completion evidence for every reviewer run; see Completion Evidence.

Agent	Description
trackbed:a11y-auditor	Audits UI code for WCAG 2.1 AA issues: semantics, ARIA, keyboard and focus handling, screen readers, and contrast.
trackbed:architecture-strategist	Reviews module design: deep modules, use of public interfaces, change locality, and whether new seams have real variation.
trackbed:cleanup-reviewer	Reviews changes for code to delete, replace with existing tools, reuse, or simplify, and checks that removals are safe.
trackbed:code-reviewer	Reviews recent code against project guidelines, CLAUDE.md conventions, and established patterns.
trackbed:codebase-pattern-reviewer	Checks a spec or design for new patterns or dependencies without rationale, and for statements about today's code that the code contradicts.
trackbed:cold-reader	Reads material as an outsider, without the conversation or project instructions, for cold reads, comprehension reads, and unprimed ratings.
trackbed:comment-analyzer	Reviews code comments for accuracy, necessity, completeness, and long-term maintainability.
trackbed:convention-drift-reviewer	Flags drift from the codebase's established practice and overly defensive code, citing peer-file evidence.
trackbed:copy-reviewer	Flags UI text that merely restates what the interface already communicates.
trackbed:design-iterator	Refines a design over several screenshot-review rounds when it still feels off.
trackbed:deslop-reviewer	Detects AI-generated slop in code changes, or in review suggestions that would introduce it.
trackbed:overengineering-reviewer	Asks whether a change is overdone, needlessly complex, or hedging against unlikely edge cases, then defends or confirms each finding.
trackbed:performance-oracle	Reviews code for performance and scalability risks: complexity, queries, memory, caching, and contention.
trackbed:pr-relevance-validator	Filters review findings down to those the current changes caused, dropping pre-existing issues.
trackbed:pr-test-analyzer	Reviews a PR's test coverage for untested behavior, edge cases, and regressions.
trackbed:product-reviewer	Reviews PRs, specs, or live flows from a product perspective: discoverability, flows, edge cases, copy, and fit.
trackbed:security-reviewer	Reviews changes for exploitable flaws: injection, authentication and access control, secrets and crypto, and business-logic bugs.
trackbed:silent-failure-hunter	Finds swallowed errors, weak error propagation, and misleading fallbacks.
trackbed:type-design-analyzer	Reviews new or changed types for encapsulation, invariants, usefulness, and enforcement.
trackbed:ux-reviewer	Reviews changes for usability with Nielsen's heuristics: missing states, confusing flows, poor feedback, and weak recovery.
trackbed:visual-reviewer	Reviews visual consistency and responsive behavior: design tokens, spacing, typography, color, and breakpoints.