review

dev360 Claude Plugins

Install

# Add marketplace
claude plugin marketplace add dev360/claude

# Install plugins
claude plugin install dev@dev360 --scope user
claude plugin install review@dev360 --scope user
claude plugin install debugger@dev360 --scope user
claude plugin install plan@dev360 --scope user

Update

# Pull latest from marketplace
claude plugin marketplace update dev360

# Update individual plugins
claude plugin update dev@dev360 --scope user
claude plugin update review@dev360 --scope user
claude plugin update debugger@dev360 --scope user
claude plugin update plan@dev360 --scope user

Restart Claude after updating.

Plugins

dev

Investigation and self-improvement agents.

• triage - Automated bug/incident investigation. Parses tickets, gathers evidence from logs and session replays, traces symptoms to code. Stops for human verification at key checkpoints.
• retrospective - Post-mortem agent that runs after painful tasks to improve docs and prompts. Only triggers when corrections were needed or effort was wasted.

review

Code review via 9 parallel specialized agents. Invoked with /review.

Agents are selected based on change size:

• Small changes (1-50 lines): logic, boundary, error-handling, security
• Medium changes (51-300 lines): adds data-flow, contracts, test-gaps
• Large changes (300+ lines): adds idioms, architecture

Each agent hunts for specific issues — logic errors, edge cases, missing error handling, security holes, blast radius, breaking API changes, test gaps, non-idiomatic patterns, and code smells. Findings are consolidated with pass/warn/fail verdicts per agent.

plan

Plan review via 9 parallel specialized agents. Invoked with /plan. Tailored for product engineers.

Agents are selected based on plan scope:

• Task (ticket, bug fix, Figma handoff): problem, prior-art, technical
• Feature (multi-file, user-facing): adds user-journey, ui-ux, delivery
• Epic (new product area, large capability): adds strategy, information-architecture, reusability

Hunts for vague requirements, missing user flows, unspecified interaction states, codebase pattern conflicts, PLG gaps, navigation dead-ends, premature abstractions, delivery risks, and missing rollback strategies. Produces a Definition of Ready checklist and pass/warn/fail verdicts per agent.

debugger

Browser debugging via Chrome DevTools MCP. Launches a subagent that can navigate pages, click elements, fill forms, monitor network requests, capture console errors, and take screenshots. Used automatically when Claude needs eyes in the browser, or explicitly via /debug.