development

Designs implementation for one epic or feature from a PRD — produces file paths, acceptance criteria, verification commands, and dependency analysis for AI-ready tickets. <example> Context: /tickets assigns each epic to a separate code-architect agent in parallel user: Create tickets for the auth login sequence epic agent: Reads the PRD section and codebase context, then produces ticket-sized units with file paths, acceptance criteria, verification commands, and dependency links </example> <example> Context: /tickets needs detailed implementation design for a complex feature user: Design the environment checking system from the PRD agent: Identifies natural ticket boundaries, maps file dependencies, and produces S/M/L complexity estimates for each unit of work </example>

Reviews PR diffs for bugs, logic errors, missing error handling, and pattern violations. Always runs on every review. <example> Context: /pr-review always dispatches this agent as the core quality gate user: Review PR #42 agent: Scans the diff for null access, race conditions, missing error handling, and pattern violations, reporting each with file:line evidence and fix suggestions </example> <example> Context: A PR adds async code with potential unhandled rejections user: Review this PR that adds the install orchestrator agent: Finds a fire-and-forget async call missing await and an overly broad catch block, reports both with impact analysis </example>

Finds opportunities to simplify code without changing behavior. Reduces nesting, eliminates redundancy, improves readability. Always runs, in parallel with other reviewers. <example> Context: /pr-review dispatches this agent in the parallel specialist batch user: Review PR #42 agent: Finds a 4-level nested conditional that can be flattened with early returns, and a wrapper function that just passes arguments through unchanged </example> <example> Context: A PR adds verbose async handling that could be simplified user: Review this PR that adds the auth login sequence agent: Identifies callback nesting that could use async/await and three duplicate error formatting blocks that should be a shared helper </example>

Explores a specific aspect of the project's codebase — architecture, patterns, or integration points. Reports findings back to the main model for synthesis. <example> Context: /tickets re-explores the codebase after reading the PRD to gather fresh implementation context user: Create tickets from the latest PRD agent: Explores relevant modules and patterns so code-architect agents have accurate codebase context for designing tickets </example> <example> Context: /tickets needs to verify that PRD assumptions match the actual codebase state user: The PRD references a notifications module — does it exist yet? agent: Searches for the referenced module and reports whether it exists yet, plus related patterns </example> <example> Context: /plan needs to understand how the project is structured before writing a PRD user: Plan a new notifications feature agent: Explores module boundaries, entry points, and existing UI/data patterns in the codebase, then reports file paths and architectural findings </example> <example> Context: /plan dispatches multiple explorers in parallel to cover architecture, patterns, and integration user: I want to add analytics tracking agent: Maps the dependency graph, identifies where tracking hooks would plug in, and reports integration constraints </example>

Checks whether code comments match the actual code. Finds comment rot, misleading docs, and outdated explanations. Runs when changed files contain code comments. <example> Context: /pr-review detects that changed files contain JSDoc or inline comments user: Review PR #42 agent: Compares each comment in the diff against the actual code behavior, flagging stale parameter descriptions and misleading algorithm explanations </example> <example> Context: A function was refactored but its doc comment still describes the old behavior user: Review this PR that refactors the auth flow agent: Finds that the @returns annotation says 'token string' but the function now returns a Result object, and flags the mismatch </example>

Implement GitHub Issue tickets sequentially with subagent execution, lightweight review, and PR creation. Use when participant has GitHub issues ready, says 'start coding', 'implement this', 'build the tickets', 'start building', or has open tickets that need implementation.

Diagnose where you are in the workshop and get back on track. Use when participant seems stuck, confused, says 'I'm lost', 'help me', 'where am I', 'what do I do next', 'I'm stuck', 'what happened', or asks about workshop flow. Also use when Claude detects the participant is struggling or has gone off track.

Deploy your project to Vercel — get a live URL anyone can visit. Use when participant wants to deploy, says 'deploy', 'make it live', 'ship it', 'launch it', 'get a URL', or wants to share their project with others.

Review a PR with specialist agents and confidence scoring — surfaces only high-confidence findings. Use when participant has a PR ready, says 'review my code', 'check this PR', 'is this ready', 'code review', or has an open pull request that needs specialist review.

Create a PRD through guided discovery, codebase exploration, and architecture discussion. Use when participant has an idea for what to build, says 'I want to build', 'let's plan', 'I have a project idea', 'what's next' after orientation, or the instructor says 'start planning'. Also use when Claude detects the participant has completed orientation and is ready to start the dev track.