Production-grade engineering workflow plugin for Claude Code. End-to-end feature pipeline, 8 specialized agents, quality gates, isolated worktrees, and persistent memory. Ship features, not prompts.
Agent
System design and architecture agent. Use for: designing new systems, evaluating tradeoffs, planning migrations, reviewing architecture, API design, database schema design. Advisory only — does not modify code.
Agent
Code-level reviewer. Use for: checking whether a change does the thing RIGHT — bugs, edge cases, security, performance, readability, and pattern violations in the implementation. Catches 'thing wrong' errors. Read-only.
Agent
Convention review lens. Use for: checking naming, file structure, and adherence to the project's existing conventions and patterns in a change. Read-only — analyzes and reports, cannot modify code.
Agent
Root-cause debugging agent that finds and fixes bugs in an isolated worktree. Use for: diagnosing failures, tracking down root causes, fixing regressions. Returns a branch with a minimal, tested fix and all quality gates green.
Agent
Documentation agent. Use for: generating JSDoc/docstrings, writing READMEs, API documentation, architecture decisions, inline comments. Follows existing patterns.
Agent
Use when /build N spawns per-step implementation: implement one PLAN.md step in an isolated worktree, run the quality gates, and make one atomic commit. Brief-driven and gate-aware — unlike @writer (general-purpose), this agent works a single plan step against its brief and stops.
Agent
Adversarial skeptic that tries to REFUTE a single recorded review finding against the unit diff; returns UPHELD / REFUTED-not-real / REFUTED-over-rated. Read-only.
Agent
Incident responder that reconstructs the timeline, assesses blast radius, and recommends remediation. Use for: production incidents, outages, regressions in the wild, post-incident analysis. Read-only — analyzes and reports, does not apply fixes.
Agent
Migration specialist that performs schema, dependency, and framework migrations in an isolated worktree. Use for: database schema changes, dependency upgrades, framework version bumps, API contract migrations. Returns a branch with a reversible, incrementally-tested migration.
Agent
Performance engineer that profiles code, identifies bottlenecks, and recommends optimizations. Use for: latency/throughput analysis, hotspot identification, algorithmic complexity review, memory/allocation analysis. Read-only — analyzes and reports, does not apply changes.
Agent
Performance review lens. Use for: spotting algorithmic complexity, wasteful allocations, N+1 queries, and hot-path regressions in a change. Read-only — analyzes and reports, cannot modify code.
Agent
Tactical planning agent. Use this agent when a brief exists and you need ordered, dependency-aware steps plus a verification plan before building — i.e. when the user runs /plan N or asks to break a brief into a buildable plan. Reads the brief, writes one PLAN.md; does not write source code.
Agent
Deep codebase and web research agent. Use for: finding patterns, tracing data flows, understanding how things work, gathering context before implementation. Spawned for parallel research tasks.
Agent
Security audit agent. Use for: vulnerability scanning, security code review, dependency auditing, secrets detection, auth/authz analysis, OWASP compliance. Read-only.
Agent
Spec-level reviewer. Use for: checking whether a change does the RIGHT thing — that it matches its brief, spec, and acceptance criteria. Catches scope gaps, missing requirements, and 'built the wrong thing' errors. Read-only.
Agent
Test review lens. Use for: checking test coverage, test quality, and missing cases or edge cases for a change. Read-only — analyzes and reports, cannot modify code.
Agent
Test generation agent. Use for: adding test coverage to existing code, writing comprehensive test suites, covering edge cases. For TDD on new features, use /tdd skill instead.
Agent
Goal-backward verification agent. Use this agent when a unit has been built and you need to know what is truly done — i.e. when the user runs /verify N or asks whether the brief's goals are met. Reads the brief's acceptance criteria, gathers concrete evidence, and classifies each criterion COVERED / PARTIAL / MISSING.…
Agent
Implementation agent that writes production-grade code in an isolated worktree. Use for: implementing features, fixing bugs, building components. Returns a branch with verified, tested, quality-gated changes.