sources: brewcode/skills/*/SKILL.md, brewcode/.claude-plugin/plugin.json focus: Commands (table must contain ALL skills from skills/), version (from plugin.json) preserve: ## Installation, ## Quick Start checks: - Compare command list with `ls skills/` — all must be in the table - Version in ## Version must match plugin.json - Links to docs/*.md must exist

Brewcode

A plugin for Claude Code that executes tasks of any size through automatic state handoff between context compactions. Creates a specification, a phased plan, launches execution with multi-agent verification, and accumulates knowledge throughout the entire task.

v3 Architecture: Task API-driven execution. Phase details live in individual phases/ files -- the manager stays slim, agents read only their assigned phase. Parallel execution via Task API groups. Lighter coordinator focused on knowledge extraction and report verification.

Installation

From plugin directory

claude --plugin-dir ./brewcode

From marketplace

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

Quick Start

/brewcode:setup                              # 1. Adapt templates for the project
/brewcode:spec "Implement JWT authorization"  # 2. Research + specification
/brewcode:plan                                # 3. Generate phased plan
/brewcode:start                               # 4. Execute with infinite context

After /brewcode:setup, templates are adapted once. Then for each task -- the cycle spec - plan - start.

v3 Flow

/brewcode:plan  --> Slim PLAN.md (Phase Registry table)
                --> phases/ directory with individual phase files
                    (1-research.md, 1V-verify-research.md, 2-implement.md, ...)

/brewcode:start --> Manager reads PLAN.md Phase Registry only
                --> TaskCreate spawns agent with phase file path
                --> Agent reads phases/{N}-{name}.md
                --> Parallel group phases spawn simultaneously
                --> Coordinator: knowledge extraction + report verification

Commands

Command	Description
`/brewcode:setup`	Project analysis, generation of adapted templates and configuration
`/brewcode:spec`	Codebase research, user dialogue, SPEC.md creation
`/brewcode:plan`	Execution plan generation from SPEC or Plan Mode with quorum review
`/brewcode:start`	Task launch with infinite context through automatic handoffs
`/brewcode:teams`	Dynamic agent team creation, management, and tracking
`/brewcode:rules`	Rule extraction from accumulated knowledge to `.claude/rules/`
`/brewcode:grepai`	Semantic code search (setup, status, start, stop, reindex)
~~`/brewcode:text-optimize`~~	moved to brewtools
~~`/brewcode:text-human`~~	moved to brewtools
~~`/brewcode:secrets-scan`~~	moved to brewtools
`/brewcode:agents`	Interactive agent creation and improvement
`/brewcode:skills`	Skill management and activation
`/brewcode:standards-review`	Standards compliance review
`/brewcode:convention`	Extract etalon classes, patterns, architecture into convention docs and rules
`/brewcode:glm-design-to-code`	GLM vision design-to-code: screenshot → multi-framework code, review, fix
`/brewcode:teardown`	Plugin configuration cleanup (tasks are preserved)

Note: /brewcode:review -- local skill, created in the project during /brewcode:setup.

Moved to brewtools: text-optimize, text-human, secrets-scan skills and text-optimizer agent are now in the brewtools plugin. Install brewtools and use /brewtools:text-optimize, /brewtools:text-human, /brewtools:secrets-scan.

Moved to brewdoc: /brewcode:auto-sync is now in the dedicated brewdoc plugin. Install brewdoc and use /brewdoc:auto-sync.

Detailed description of each command: docs/commands.md

Agents

Specialized agents spawned by brewcode skills during task execution:

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

Agent	Purpose	Best for
developer	Full-stack development -- implements features, writes code, fixes bugs	Implementation, bug fixes, refactoring, unit tests
reviewer	System architect and code reviewer -- architecture, quality, security, performance	Code review, architecture analysis, SOLID enforcement
tester	SDET/QA -- runs tests, analyzes results, debugs flaky tests	Test execution, failure analysis, test infrastructure
architect	Architecture analysis -- design, patterns, trade-offs, scaling strategies	Architecture review, module decomposition, pattern evaluation
~~text-optimizer~~	moved to brewtools	Prompt compression, CLAUDE.md optimization, verbose docs
bash-expert	Creates production-quality bash/sh scripts for macOS and Linux	Shell scripts, install scripts, plugin automation
skill-creator	Creates and improves Claude Code skills (SKILL.md files)	New skill creation, skill invocation debugging
agent-creator	Creates and improves Claude Code agents	New agent creation, agent triggering improvements
hook-creator	Creates and debugs Claude Code hooks (lifecycle event handlers)	Hook creation, hook debugging, schema validation
bc-coordinator	Task coordinator -- knowledge extraction, report verification, FINAL.md	Phase tracking, knowledge management, task finalization
bc-knowledge-manager	KNOWLEDGE.jsonl compaction -- deduplication, prioritization, truncation	Knowledge cleanup before handoff, duplicate removal
bc-grepai-configurator	grepai config specialist -- project analysis, config.yaml generation	Semantic search setup, grepai configuration
bc-rules-organizer	Creates and optimizes `.claude/rules/*.md` with path-specific frontmatter	Rule extraction, CLAUDE.md splitting, rule organization

Configuration

Configuration file is created during /brewcode:setup:

.claude/tasks/cfg/brewcode.config.json

Main sections:

Section	Purpose
`knowledge`	Entry limits, validation, retention (global/task)
`constraints`	Role constraints for agents (DEV, TEST, REVIEW)
`autoSync`	Synchronization interval, parallelism

Task Structure

After task creation, the following appears in the project:

.claude/tasks/{TS}_{NAME}_task/
  SPEC.md             # Specification
  PLAN.md             # Slim plan: 3-line header + Phase Registry table
  KNOWLEDGE.jsonl     # Accumulated knowledge
  phases/             # Individual phase files for agents (v3)
    1-research.md
    1V-verify-research.md
    2-implement.md
    FR-final-review.md
  artifacts/          # Execution reports by phases
  backup/             # Backups

Documentation

Document	Description
`docs/commands.md`	Detailed description of all commands with examples
`docs/hooks.md`	Hooks and their behavior
`docs/flow.md`	Execution flow diagrams (spec, plan, start)
`docs/file-tree.md`	Complete file structure of plugin and project
INSTALL.md	Installation guide
RELEASE-NOTES.md	Change history

Version

3.4.22 -- latest release.

Author: Maksim Kochetkov | License: MIT