Show HN: Claude Code plugin that runs a semi-autonomous AI dev workflow (github.com)

Claude Code Workflows is a Claude Code plugin that provides a semi‑autonomous, opinionated development workflow: install the plugin in Claude Code, then run high‑level commands like /implement, /task, /design, /plan, /build and /review to trigger a chain of specialized agents that handle requirements analysis, PRD/design creation, task decomposition, TDD implementation, automated fixes, testing and code review. The package ships 11 purpose‑built agents (requirement‑analyzer, prd‑creator, technical‑designer, work‑planner, task‑decomposer, task‑executor, quality‑fixer, code‑reviewer, acceptance‑test‑generator, rule‑advisor, document‑reviewer), a rules directory for coding/testing/documentation standards, and templates to enforce consistent, production‑ready outputs. It also documents practical setup steps (e.g., SSH key setup for GitHub) and exposes agent/rule customization under ~/.claude/agents/. For AI/ML teams this matters because it operationalizes best practices (TDD, coverage, architecture patterns) into repeatable, language‑agnostic agents, reducing context exhaustion and quality drift in long sessions while accelerating end‑to‑end delivery. Technically notable aspects: agent specialization to preserve fresh context per phase, automated QA and type/tests generation, commit‑sized task decomposition, and configurable rule sets so teams can tune tradeoffs. Early benchmarks claim multi‑day feature delivery (e.g., 30 TypeScript files with full coverage in ~2 days). Practical caveats include trusting auto‑fixes and CI integration/security (SSH/Git access) when adopting an autonomous workflow.