Brewcode

Infinite task execution plugin for Claude Code -- automatic context handoff, multi-agent workflows, knowledge persistence.

Field	Value
Version	4.4.0
Skills	9
Agents	10
Hooks	2
Model	opus

Install

Paste this into a Claude Code session:

Execute these commands in this session, one by one, show full output for each, do not skip any:

1. claude plugin marketplace add https://github.com/kochetkov-ma/claude-brewcode
2. claude plugin install brewcode@claude-brewcode

After install, run `/reload-plugins` (or `exit` + `claude`).

Or install the whole suite

Execute these commands in this Claude Code session, one by one, show full output for each, do not skip any, do not summarize:

1. claude plugin marketplace add https://github.com/kochetkov-ma/claude-brewcode
2. claude plugin install brewcode@claude-brewcode
3. claude plugin install brewdoc@claude-brewcode
4. claude plugin install brewtools@claude-brewcode
5. claude plugin install brewui@claude-brewcode

After all commands succeed, run `/reload-plugins`. If `/reload-plugins` is unavailable, tell me to type `exit` and run `claude` again. Run the commands now.

Update anytime with /brewtools:plugin-update.

Overview

Brewcode turns single Claude Code sessions into an infinite task pipeline. Claude Code's native auto-compaction preserves the working context, and brewcode hooks re-inject plugin state on each session so the task runs to completion regardless of how many compaction cycles occur.

Skills cover project analysis, specification creation through parallel research agents, code review, convention analysis, and project rules management. Specialized agents handle implementation, testing, review, architecture, and coordination.

Installation

# Marketplace (recommended)
claude plugin marketplace add https://github.com/kochetkov-ma/claude-brewcode
claude plugin install brewcode@claude-brewcode

# Already installed? Update
claude plugin marketplace update claude-brewcode
claude plugin update brewcode@claude-brewcode

# Dev mode (no install)
claude --plugin-dir ./brewcode

Quick Start

/brewcode:spec "Implement JWT authorization"  # Research + specification

Skills

Skill	Purpose
`/brewcode:spec`	Research codebase + user dialog -> SPEC.md
`/brewcode:superreview`	Generate a project-tailored deep-review skill: domain-expert routing + scope discipline + mechanical gates + adversarial validation
`/brewcode:teams`	Dynamic agent team creation, management, and performance tracking
`/brewcode:convention`	Extract etalon classes, patterns, architecture into convention docs and rules
`/brewcode:rules`	Prompt-driven rules management: status, create, improve, review
`/brewcode:skills`	Prompt-driven skill management: status, create, improve, sync, review
`/brewcode:agents`	Prompt-driven agent management: status, create, improve, sync, review
`/brewcode:e2e`	E2E testing orchestration with BDD scenarios and quorum review

Note: /brewcode:superreview emits a self-contained, project-local deep-review skill tailored to your stack. It first makes sure the project HAS domain experts (creating the missing ones via agent-creator), then wires the emitted skill to the project's real gates, rules and scope baseline (task + issue + recorded decisions).

Agents

Agent	Model	Purpose
developer	inherit	Implements features, writes code, fixes bugs, refactors
tester	sonnet	Runs tests, analyzes failures, debugs flaky tests
reviewer	inherit	Reviews architecture, quality, security, performance
architect	inherit	Architecture analysis, patterns, scaling
skill-creator	inherit	Creates and improves Claude Code skills
agent-creator	inherit	Creates and improves Claude Code agents
hook-creator	inherit	Creates and debugs Claude Code hooks
bash-expert	inherit	Creates sh/bash scripts for Mac/Linux
bc-rules-organizer	haiku	Internal: spawned by /brewcode:rules

Dynamic teams: Use /brewcode:teams create to generate 5-20 project-specific agents with self-selection protocol and performance tracking.

Scope guard: every agent carries a ## Scope guard -- if a task exceeds one bounded unit (one deliverable, ~5 files), the agent stops and proposes a split instead of running for an hour.

Architecture

brewcode/
+-- .claude-plugin/plugin.json          # Plugin manifest
+-- hooks/                              # 2 lifecycle hooks
|   +-- session-start.mjs              # SessionStart: version-check, plan-symlink, permission_mode
|   +-- forced-eval.mjs                # UserPromptSubmit: manager-role + split-discipline reminder
|   +-- hooks.json                     # Event bindings
|   +-- lib/utils.mjs                  # Shared utilities
+-- agents/                            # 9 agents
+-- skills/                            # 8 skills
+-- templates/                         # Rule templates

Hook Lifecycle

Hook	Event	Purpose
session-start	SessionStart	Version-check, plan-symlink, permission_mode tag
forced-eval	UserPromptSubmit	Manager-role + split-discipline reminder, 2 lines via additionalContext (9K bound)

Task Structure

.claude/tasks/{TS}_{NAME}_task/
  SPEC.md             # Specification (research results from /brewcode:spec)

Documentation

Full docs: doc-claude.brewcode.app/brewcode/overview

Resource	Link
Skills reference	Skills
Agents reference	Agents
Hooks reference	Hooks
Workflow	Workflow
Release Notes	RELEASE-NOTES.md

Author: Maksim Kochetkov | License: MIT

6.4 KiB Raw Blame History