claude-skills

Use this agent to review a complete feature's code holistically after all deliverables are implemented. Reads all .claude/rules/ conventions, checks the full diff for cross-deliverable concerns (naming consistency, duplication, integration gaps), and produces a structured report. Dispatched by execute-tasks after all deliverables pass per-deliverable reviews. Examples: <example>Context: execute-tasks has completed all deliverables and per-deliverable reviews passed. user: "Run holistic code review for the reservations feature" assistant: "Dispatching the code-reviewer agent with all changed files and convention rules" <commentary>The code reviewer checks the FULL feature diff, catching cross-deliverable issues that per-deliverable quality-reviewer cannot see.</commentary></example>

Final holistic DESIGN audit of the whole feature's UI after code review. Runs the refactoring-ui-reviewer skill across every visual component, outputs prioritized findings (🔴 Critical / 🟡 Important / 🟢 Nitpick) with rule citations + file:line. Pairs with a fix-all loop. Mirrors code-reviewer/test-reviewer but for visual design.

Use this agent to implement a single task or deliverable from an implementation plan. The agent reads project conventions from .claude/rules/, implements the spec, writes tests, self-reviews, and reports status. Examples: <example>Context: The execute-tasks skill is dispatching tasks from an implementation plan. user: "Implement D3: Create the reservation card component with hook" assistant: "Dispatching the implementer agent with the full deliverable spec and relevant convention files" <commentary>The implementer agent receives one isolated task with conventions injected, implements it, and reports back.</commentary></example> <example>Context: User wants a specific piece of code built following project conventions. user: "Build the extend reservation dialog following the spec in the PRD" assistant: "Let me dispatch the implementer agent with the spec and your project conventions" <commentary>Can be used standalone outside the pipeline for any implementation task that needs convention compliance.</commentary></example>

Use this agent to review code against project conventions from .claude/rules/. Reads convention files dynamically, checks code quality and architecture patterns, tags findings as TRIVIAL (auto-fixable) or ARCHITECTURAL (needs human decision). Examples: <example>Context: The execute-tasks skill dispatches this after spec compliance passes. user: "Review code quality for D3" assistant: "Dispatching the quality-reviewer agent with the changed files and relevant convention rules" <commentary>The quality reviewer reads the convention files, then checks every line of changed code against them.</commentary></example> <example>Context: User wants a convention audit on code they wrote. user: "Check if my new hook follows our conventions" assistant: "Let me dispatch the quality reviewer to check your code against the project rules" <commentary>Can be used standalone for any convention compliance check. Reads .claude/rules/ dynamically.</commentary></example>

Use this agent to verify that an implementation matches its specification. Reads the actual code (never trusts the implementer's report), produces a compliance matrix, and flags missing requirements or scope creep. Examples: <example>Context: An implementer agent just completed a task and the execute-tasks skill needs to verify compliance. user: "Review spec compliance for D3: reservation card component" assistant: "Dispatching the spec-reviewer agent to verify the implementation matches the deliverable spec" <commentary>The spec reviewer independently reads the code and compares it line by line against the spec requirements.</commentary></example> <example>Context: User manually implemented something and wants to check it matches the PRD. user: "Check if what I built matches the PRD requirements" assistant: "Let me dispatch the spec reviewer to compare your implementation against the PRD" <commentary>Can be used standalone to verify any implementation against any spec document.</commentary></example>

Finalize a feature — run tests, build, and present options for committing, creating a PR, or discarding. Use when user says "finish", "wrap up", "done", "create PR", "merge", or wants to finalize their work on a feature.

Write tests for TruckBays features using Vitest, React Testing Library, and MSW v2. Use when writing tests, adding test coverage, or setting up testing infrastructure.

Scaffold shared infrastructure projects — providers, hooks, layout components, i18n, config/theme. Use when the project builds reusable plumbing in shared/, or when user says "scaffold shared", "create provider", "setup layout".

Execute an implementation plan using autonomous subagents. Reads the plan from plan-implementation, dispatches implementer + reviewer agents per deliverable, handles smart triage. Use when user says "execute the plan", "build it", "start implementing", or after plan-implementation completes.

Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.