Plugin

maestro

Name: maestro
Author: jbj338033

Thoroughness-first orchestrator — verification gates, multi-model teams, context persistence

npx claudepluginhub jbj338033/maestro --plugin maestro

Component Overview

Agents

Skills

Hooks

MCP Servers

Component Details

Agents (7)

codex-bridge

/codex-bridge

Bridge to OpenAI Codex CLI. Delegates tasks to Codex and returns structured results. Use for code implementation, refactoring, debugging, or getting a second opinion from a different model.

critic

/critic

Deep multi-perspective code review. Security, performance, maintainability analysis. Use for significant changes (5+ files) or architectural decisions. Opus-powered.

gemini-bridge

/gemini-bridge

Bridge to Google Gemini CLI. Delegates tasks to Gemini and returns structured results. Strong at math, logic, long-context analysis, and multimodal reasoning.

healer

/healer

Automated error diagnosis and repair. Analyzes test/build failures, identifies root cause, and attempts targeted fixes.

researcher

/researcher

Read-only codebase exploration. Discovers patterns, conventions, and related files before implementation. Use proactively before making changes to understand existing code.

synthesizer

/synthesizer

Reconciles outputs from multiple models (Claude, Codex, Gemini). Identifies agreements, disagreements, and blind spots. Use after getting responses from multiple models.

verifier

/verifier

Evidence-based verification. Runs tests, checks types, verifies build. Every claim must have command→output evidence. Use after implementation to confirm work is correct.

Skills (3)

ask

/ask

Ask Codex or Gemini a specific question

setup

/setup

Set up Maestro — install HUD wrapper, register statusline, verify environment

status

/status

Show current Maestro session state, mission progress, and memory

Hooks (1)

Review workflow modifications before installing

Event Hooks

All tools

11 hooks across 9 events

MCP Servers (1)

Connects to external services

maestro

admin

README

`maestro`

Intelligent harness for Claude Code
Amplify, guard, and learn — automatically.

Install · How it works · Agents · Skills · MCP Tools

Why

Claude Code is powerful but has no memory between sessions, no project awareness at start, and no structural enforcement of quality. Maestro fixes all three — automatically, with zero configuration.

AMPLIFY — Scans your project at session start and injects structure, conventions, and past decisions into context
GUARD — Blocks completion without tests, warns on dangerous patterns, detects circular edits
LEARN — Auto-captures decisions, commands, and patterns. Every session makes the next one smarter

Install

claude plugin add jbj338033/maestro

Requires Node.js >= 22 and Claude Code with plugin support.

How it works

AMPLIFY — Smarter from the first message

Feature	How
Project scan	SessionStart scans `package.json`, `Cargo.toml`, directory tree → injects language, framework, scripts, entry points
Memory injection	Past conventions and decisions are injected as actual content, not just counts
Learned commands	Test/build commands from past sessions are surfaced automatically
Directory context	When you first read a file in a directory, its README.md is auto-injected

GUARD — Mistakes caught automatically

Feature	How
Stop gate	Blocks completion if tests weren't run after file changes
Verification reset	`tests_passed` resets when new files are modified after passing
Circular edit detection	Warns after 3+ consecutive edits without re-reading
Test file reminder	Reminds you when `foo.test.ts` exists for a modified `foo.ts`
Dangerous patterns	`git add -A`, force push, `--hard`, `rm -rf`, `--no-verify`, secret exposure
Mission verification	Agent verifies acceptance criteria before allowing completion

LEARN — Gets smarter over time

Captured	When	Used
Package installs	`npm/pnpm install X`	Remembered as decisions
Test commands	`pnpm test`, `cargo test`, etc.	Injected at next session start
Build commands	`pnpm build`, `next build`, etc.	Injected at next session start
Config changes	Edits to `tsconfig.json`, etc.	Remembered as decisions
File co-modification	Which directories change together	Pattern awareness

Hooks

Hook	What it does
SessionStart	Project scan + memory injection + compact state restore
PreToolUse	Dangerous pattern warnings + directory context injection + circular edit detection
PostToolUse	File tracking + test/build detection + auto-capture decisions + test file reminders
Stop	Verification gate + cross-validation reminder + mission criteria verification
PreCompact	Save state + notes + decisions before compression
PostCompact	Restore state after compression
SubagentStart/Stop	Track active subagents
SessionEnd	Save history + learn patterns + cleanup subagents

Agents

Agent	Model	Writes?	Purpose
researcher	Sonnet	No	Read-only codebase exploration
verifier	Sonnet	No	Evidence-based verification
critic	Opus	No	Security, performance, maintainability review
codex-bridge	Sonnet	Yes	Delegates to OpenAI Codex CLI
gemini-bridge	Sonnet	Yes	Delegates to Google Gemini CLI
synthesizer	Opus	No	Reconciles multi-model outputs

Skills

Skill	Usage	Description
ask	`/maestro:ask <codex\|gemini> "question"`	Query an external model
status	`/maestro:status`	Show session state, mission progress
setup	`/maestro:setup`	Install HUD statusline, verify environment

MCP Tools

Tool	Description
`state_read`	Read session state
`state_write`	Update session state
`mission_create`	Create mission with acceptance criteria
`mission_read`	Read current mission
`mission_update`	Mark criteria verified/unverified
`memory_read`	Read cross-session memory
`memory_write`	Add to cross-session memory
`memory_search`	Search memory by keyword
`history_list`	List past session history

Architecture

View full README on GitHub

Similar Plugins

maestro

206

Elite AI Development Orchestrator for Claude Code. Provides specialized agents, modular skills, intelligent hooks, and persistent memory systems for autonomous development workflows.

Stats

Version0.1.0

Stars0

MaintenanceExcellent

AddedMar 30, 2026

Actions

View on GitHub View README Plugin Marketplace JSON

Available In

maestro

Safety Signals

Critical

Matches all tools

Hooks run on every tool call, not just specific ones

Admin access level

Server config contains admin-level keywords

Caution

Help us improve

Share bugs, ideas, or general feedback.

Back to Plugins

`maestro`

Intelligent harness for Claude Code
Amplify, guard, and learn — automatically.

Install · How it works · Agents · Skills · MCP Tools

Why

AMPLIFY — Scans your project at session start and injects structure, conventions, and past decisions into context
GUARD — Blocks completion without tests, warns on dangerous patterns, detects circular edits
LEARN — Auto-captures decisions, commands, and patterns. Every session makes the next one smarter

Install

claude plugin add jbj338033/maestro

Requires Node.js >= 22 and Claude Code with plugin support.

How it works

AMPLIFY — Smarter from the first message

Feature	How
Project scan	SessionStart scans `package.json`, `Cargo.toml`, directory tree → injects language, framework, scripts, entry points
Memory injection	Past conventions and decisions are injected as actual content, not just counts
Learned commands	Test/build commands from past sessions are surfaced automatically
Directory context	When you first read a file in a directory, its README.md is auto-injected

GUARD — Mistakes caught automatically

Feature	How
Stop gate	Blocks completion if tests weren't run after file changes
Verification reset	`tests_passed` resets when new files are modified after passing
Circular edit detection	Warns after 3+ consecutive edits without re-reading
Test file reminder	Reminds you when `foo.test.ts` exists for a modified `foo.ts`
Dangerous patterns	`git add -A`, force push, `--hard`, `rm -rf`, `--no-verify`, secret exposure
Mission verification	Agent verifies acceptance criteria before allowing completion

LEARN — Gets smarter over time

Captured	When	Used
Package installs	`npm/pnpm install X`	Remembered as decisions
Test commands	`pnpm test`, `cargo test`, etc.	Injected at next session start
Build commands	`pnpm build`, `next build`, etc.	Injected at next session start
Config changes	Edits to `tsconfig.json`, etc.	Remembered as decisions
File co-modification	Which directories change together	Pattern awareness

Hooks

Hook	What it does
SessionStart	Project scan + memory injection + compact state restore
PreToolUse	Dangerous pattern warnings + directory context injection + circular edit detection
PostToolUse	File tracking + test/build detection + auto-capture decisions + test file reminders
Stop	Verification gate + cross-validation reminder + mission criteria verification
PreCompact	Save state + notes + decisions before compression
PostCompact	Restore state after compression
SubagentStart/Stop	Track active subagents
SessionEnd	Save history + learn patterns + cleanup subagents

Agents

Agent	Model	Writes?	Purpose
researcher	Sonnet	No	Read-only codebase exploration
verifier	Sonnet	No	Evidence-based verification
critic	Opus	No	Security, performance, maintainability review
codex-bridge	Sonnet	Yes	Delegates to OpenAI Codex CLI
gemini-bridge	Sonnet	Yes	Delegates to Google Gemini CLI
synthesizer	Opus	No	Reconciles multi-model outputs

Skills

Skill	Usage	Description
ask	`/maestro:ask <codex\|gemini> "question"`	Query an external model
status	`/maestro:status`	Show session state, mission progress
setup	`/maestro:setup`	Install HUD statusline, verify environment

MCP Tools

Tool	Description
`state_read`	Read session state
`state_write`	Update session state
`mission_create`	Create mission with acceptance criteria
`mission_read`	Read current mission
`mission_update`	Mark criteria verified/unverified
`memory_read`	Read cross-session memory
`memory_write`	Add to cross-session memory
`memory_search`	Search memory by keyword
`history_list`	List past session history

maestro

Component Overview

Component Details

Agents (7)

Skills (3)

Hooks (1)

MCP Servers (1)

README

maestro

Why

Install

How it works

AMPLIFY — Smarter from the first message

GUARD — Mistakes caught automatically

LEARN — Gets smarter over time

Hooks

Agents

Skills

MCP Tools

Architecture

Similar Plugins

maestro

Help us improve

Help us improve

maestro

Component Overview

Component Details

Agents (7)

Skills (3)

Hooks (1)

MCP Servers (1)

README

maestro

Why

Install

How it works

AMPLIFY — Smarter from the first message

GUARD — Mistakes caught automatically

LEARN — Gets smarter over time

Hooks

Agents

Skills

MCP Tools

Architecture

Similar Plugins

maestro

Help us improve

claude-workflow

evolving-lite

maister

everything-claude-code

caveman

Help us improve

`maestro`

`maestro`