The Project Manager Extension for Gemini CLI – Conductor

Conductor is a free, open-source Gemini CLI extension that transforms your Gemini CLI into an AI project manager. It enforces a structured development workflow where you create specifications and plans before writing code. This shifts project context out of ephemeral chat windows and into persistent Markdown files that live alongside your codebase.

The extension formalizes your development intent through a Context → Spec & Plan → Implement lifecycle. You maintain complete control over what gets built because Conductor requires human approval at each stage. The extension reads your project’s architecture, style guides, and technical preferences to ensure AI agents produce code that matches your standards.

Conductor works with both new greenfield projects and existing brownfield codebases. Set up project-level context once, and your entire team can reference that shared foundation for every feature they build.

Features

• Context Management: Store project documentation, style guides, tech stack preferences, and workflow rules in Markdown files that persist across all development sessions.
• Specification Generation: Create detailed requirement documents for each feature or bug fix before any code is written.
• Plan Review System: Approve or modify action plans that break work into phases, tasks, and subtasks before implementation begins.
• Team Collaboration: Share project context across developers so AI agents maintain consistency regardless of who runs commands.
• Brownfield Support: Initialize context for existing projects through interactive sessions that capture architecture and guidelines.
• Smart Revert: Undo work at the track, phase, or task level using git-aware commands that understand logical units instead of commit hashes.

Use Cases

• Feature Development: Plan new capabilities by defining user needs, technical requirements, and implementation steps before writing code.
• Bug Fix Tracking: Document bugs with specs that explain the issue, root cause, and fix strategy.
• Team Onboarding: New developers clone your repository and inherit all project context automatically.
• Workflow Standardization: Enforce Test-Driven Development or other methodologies by defining them in workflow.md.
• Cross-Project Consistency: Maintain coding standards across multiple repositories by reusing context templates.

How to Use It

1. Install Conductor with automatic updates:

gemini extensions install https://github.com/gemini-cli-extensions/conductor --auto-update

2. Execute /conductor:setup once per project to establish core context. The command prompts you to define four categories:

• Product context covers users, goals, and high-level features
• Product guidelines set standards for prose style, brand messaging, and visual identity
• Tech stack configures languages, databases, and frameworks
• Workflow establishes team preferences like TDD or commit strategies

Conductor generates these files:

• conductor/product.md
• conductor/product-guidelines.md
• conductor/tech-stack.md
• conductor/workflow.md
• conductor/code_styleguides/
• conductor/tracks.md

3. Run /conductor:newTrack when you’re ready to build a feature or fix a bug. You can include a description:

/conductor:newTrack "Add dark mode toggle to settings page"

Conductor generates two critical documents:

• spec.md contains detailed requirements explaining what you’re building and why
• plan.md provides an actionable task list organized by phases and subtasks

The command also creates:

• conductor/tracks/<track_id>/spec.md
• conductor/tracks/<track_id>/plan.md
• conductor/tracks/<track_id>/metadata.json

Review these files before proceeding. You can edit the plan to adjust scope or approach.

4. Execute /conductor:implement after approving the plan. The AI agent works through tasks sequentially and updates status in plan.md as it completes each item.

The agent follows your defined workflow. For TDD, it writes tests first, runs them to confirm failure, implements code, then verifies tests pass.

Conductor guides you through manual verification at the end of each phase. This checkpoint ensures everything works before moving forward.

5. Check progress anytime:

/conductor:status

This displays track status and active task information.

6. Undo work if needed:

/conductor:revert

The command prompts you to select a track, phase, or task. Conductor analyzes git history to revert the appropriate changes.

7. Conductor reads your context, specifications, and plans during each interaction. This consumes more tokens than basic CLI usage, especially in large projects. Run /stats model to check the current session consumption.

Pros

• Context is king: It moves project awareness from transient chat logs into permanent, version-controlled files.
• Team consistency: Every AI-generated change aligns with your team’s standards[reference:7].
• Brownfield-friendly: It doesn’t require a greenfield project. It’s built to understand and adapt to existing code.
• Structured workflow: The spec-plan-implement cycle brings disciplined software development practices to AI-assisted coding.
• Human-in-the-loop: Manual verification steps and plan reviews prevent you from losing control of the codebase.

Cons

• Higher token usage: Reading context and planning files means each session consumes more tokens than a simple Gemini CLI chat.
• Gemini CLI dependency: It only works within the Gemini CLI ecosystem.

Related Resources

• Gemini CLI Extensions Documentation: Complete guide to installing, configuring, and building Gemini CLI extensions.
• Gemini CLI Extensions Gallery: Browse other extensions for integrating tools like Figma, Stripe, and Kubernetes with your development workflow.
• Best CLI AI Coding Agents: 7 Best CLI AI Coding Agents (Open-source).

FAQs

Q: Can I customize the workflow?
A: Yes. The /conductor:setup command generates a workflow.md file. You can edit this file to change how the agent behaves, such as enforcing Test-Driven Development (TDD) or specific code review steps.

Q: What happens if I dislike the plan the AI generates?
A: You can edit the plan.md file directly before running /conductor:implement.

Q: Can multiple team members use the same Conductor context?
A: Yes. The conductor directory and all context files live in your git repository. When teammates clone the project, they inherit the complete context. Everyone works from the same specifications, style guides, and workflow rules.

Q: What happens if I disagree with the generated plan?
A: You review plan.md after running /conductor:newTrack and before executing /conductor:implement. Edit the file to adjust tasks, change sequencing, or modify scope. Conductor follows whatever plan exists when you run the implement command.

Q: Can I revert just part of a feature?

A: Yes. The /conductor:revert command works at three levels: entire tracks, individual phases within a track, or specific tasks within a phase. Select the appropriate scope when prompted. Conductor analyzes git history to undo the correct changes.

Q: Does Conductor support custom workflow templates?
A: Yes. The workflow.md file uses a customizable template. You can modify it to match your team’s development practices like TDD, pair programming, or specific commit message formats. Conductor follows whatever workflow you define.