Agent Claude Code
Engineering-rules enforcer — runs deterministic SRP / Clean Architecture / TDD-delta checks against changed files and returns a structured JSON report. Invoked by /review, /commit, /pr, and plan-verifier Step 3.6.
Agent Claude Code
Structured test runner — detects stack, runs tests, parses per-stack output into strict JSON (pass/fail, counts, per-failure file+name+errorhead). Invoked by /test and plan-verifier Step 3.5. Never silently reports "not-run" as pass.
Agent Claude Code
This is a partial, not an agent. The installer places it above the text of every agent that lists it under compose:. It is the single routing table MB agents use to understand code before touching it; agents do not keep their own copies.
Agent Claude Code
Haiku-tier subagent — writes one Memory Bank wiki article from a community evidence pack.
Agent Claude Code
Sonnet-tier subagent — finds surprising cross-community connections, emits strict JSON edges.
Agent Claude Code
Plan execution auditor — rereads the plan, inspects git diff, validates every DoD item against real code. Invoked by /mb verify; REQUIRED before /mb done when work followed a plan.
Command Claude Code
A command that records important, difficult-to-reverse technical choices as Architecture Decision Records, or ADRs.
Command Claude Code
A command for maintaining a central list of confirmed project decisions, called agreements.
Command Claude Code
Runs auto-classify on the goal + diff scope: picks one route and writes it into the mb-flow fence. Use when starting a task and the route (quick fix, plan or spec) must be chosen automatically.
Command Claude Code
A command for managing API contracts: written agreements describing how software services communicate. It covers OpenAPI, gRPC, and GraphQL APIs, and can generate, check, or test them.
Command Claude Code
A command that turns a rough request and attached documents into a checked, one-page project brief saved under `briefs/`.
Command Claude Code
A command that rebuilds an agent's working context from project notes and recent Git activity.
Command Claude Code
A command that turns Git commit messages into a changelog, a record of changes between software releases.
Command Claude Code
A command that checks staged Git changes and creates a commit from them.
Command Claude Code
A command for managing `pipeline.yaml`, a version-controlled file that defines how coding work is carried out and checked.
Command Claude Code
Contract-First — defines the interface, writes contract tests, then implements. Use when building a new component or port and the abstraction must come before the code.
Command Claude Code
A database-change command that creates and manages migration files for common database tools and programming languages.
Command Claude Code
A command that interviews you in five phases and saves a detailed description of a proposed feature or change. EARS is a structured way to write requirements so they can be checked clearly.
Command Claude Code
A command for creating or updating Markdown documentation for a software module. It records the module's purpose, public functions and types, examples, dependencies, and configuration.
Command Claude Code
A session-ending command for a project that uses a memory bank, a set of files storing project context and progress. It runs required checks, updates the core memory-bank files and adds a progress note.
Command Claude Code needs its repo
A command that repeatedly follows a goal file through a fixed decision loop until the work is complete or a stop condition is reached.
Command Claude Code
Selects a route directly (skip classification) — the explicit override escape-hatch. Use when the user already knows the route and wants to bypass /mb analyze-task.
Command Claude Code
A command that creates and checks two durable project files: `goal.md` for the desired outcome and acceptance criteria, and `project.md` for lasting constraints.
Command Claude Code
A guided conversation for examining a task or idea before, during, or after implementation. It questions whether the work is needed, explores the approach, finds missing details, and records decisions.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: