workbench

Plugin workbench from marketplace workbench-ai in rafael-ventura/workbench-ai. Commands read from the repo's manifest — 26 days ago.

Skills (17)

• adr

  Record an architecture decision as a short ADR in docs/decisions - context, options considered, decision, consequences. Use when a significant technical choice is made or debated (database, messaging pattern, framework, service boundary), or when the user asks to document why something is the way it

• debug

  Systematic debugging - reproduce, gather evidence, form and test one hypothesis at a time, fix the root cause and lock it with a regression test. Use when something fails, throws, returns wrong data or behaves differently than expected, before proposing any fix.

• develop

  Entry point for building something. Sizes the request and either implements a small change directly with verification, or routes a larger feature through spec, plan, tasks and implement. Use when the user asks to build, add, change or fix something and hasn't picked a specific flow.

• git

  Git and GitHub work done safely - branches, checkout, commits, rebases and merges, pull requests and merge conflict resolution. Use when the user asks to commit, open a PR, sync with main, merge, switch branches or fix conflicts.

• implement

  Execute specs/NNN-slug/tasks.md one task at a time, verifying and ticking each box, then review the result. Use after /workbench:tasks is approved, or when the user says to implement or continue a spec.

• incident

  Investigate a production incident like a war room - build a timeline from logs, timestamps, data and deploys, separate mitigation from root cause, and write a blameless postmortem. Use when the user reports something broken in production, an outage, data inconsistency or a spike of errors.

• learn

  Teach a technology or concept by mapping it to what the user already knows (C#/.NET, Angular, Vue 3), using the user's own codebase as material, with exercises the user solves alone. Use when the user wants to learn or understand a framework, library, language feature or pattern, rather than get a t

• plan

  Turn an approved specs/NNN-slug/spec.md into plan.md, the technical approach (files, contracts, data, risks, test strategy). Use after /workbench:spec is approved, or when the user asks how to implement a spec.

• read-docs

  Read and extract what matters from PDFs and images for development work - specs, contracts, standards like ISO 20022, vendor API manuals, screenshots of errors, UI mockups and architecture diagrams - with page or region references. Use when the user shares or points to a PDF, image, screenshot, scan

• refactor

  Change the structure of code without changing its behavior, in small safe steps protected by tests - extract, rename, move, split classes, remove duplication, untangle dependency injection. Use when the user asks to refactor, clean up, simplify or reorganize code.

• review

  Code review of the current diff, a branch or a GitHub PR, focused on bugs that would break in production, then security, then simplicity. Use when the user asks for a review, says "revisa", or before opening or merging a PR.

• spec

  Start a feature the spec-driven way by writing specs/NNN-slug/spec.md (problem, requirements, acceptance criteria, out of scope, open questions) before any code. Use when the user wants to start a new feature, says "spec", or describes something non-trivial to build.

• sql-tune

  Diagnose and speed up slow SQL Server or PostgreSQL queries, stored procedures and ORM-generated SQL (EF Core, NHibernate) using execution plans, indexes and query rewrites, measuring before and after. Use when a query, report, screen or endpoint is slow because of the database, or when reviewing a

• tasks

  Break an approved specs/NNN-slug/plan.md into tasks.md, a checklist of small, ordered, verifiable steps. Use after /workbench:plan is approved.

• trace

  Understand how a flow works end to end - from the entry point (endpoint, UI action, job, message) through every layer, message and database call - and explain it with a diagram. Use when the user asks how something works, where something happens, what calls what, or before changing or debugging unfa

• unit-tests

  Analyze a class, method or change and write unit tests that protect real behavior - business rules, edge cases and past bugs - instead of chasing coverage. Use when the user asks for tests, says a piece of code is untested, or after a bug fix that needs a regression test.

• upgrade-dotnet

  Plan and execute .NET upgrades - .NET Framework to modern .NET, or between .NET versions - and NuGet package updates, project by project with breaking-change checks and green builds at every step. Use when the user wants to upgrade a target framework, migrate off .NET Framework, update packages or f

Subagents (2)

• dotnet-bug-hunter

  Analisa bugs em projetos .NET/C#. Use quando houver exception, teste falhando, comportamento inesperado em serviço, ou erro de compilação para diagnosticar. Especializado em MassTransit, EF Core, ASP.NET Core e arquitetura de microsserviços.

• messaging-reviewer

  Reviews .NET changes that publish or consume messages (MassTransit, RabbitMQ, IBM MQ, Azure Service Bus) for reliability bugs - duplicates, lost messages, retries, poison messages, ordering and saga compensation. Use after implementing or changing consumers, publishers, sagas or outbox code.

Hooks (2)

• PreToolUse

  node "${CLAUDE_PLUGIN_ROOT}/hooks/guard.mjs"

• PostToolUse

  node "${CLAUDE_PLUGIN_ROOT}/hooks/after-edit.mjs"

Similar plugins