AI Toolkit Guide

Install, configure, and master the AI Toolkit across Claude, Gemini, Codex, OpenCode, and Cursor. Use the navigation or cards below to jump directly to the instructions you need.

What is inside the toolkit?

Agents

Claude personas for planning, reviews, framework work, documentation, coaching, and reliability.

  • Files in AI-toolkit/agents/
  • See INDEX.md for full catalog
  • Copy into .claude/agents/

Prompts

Reusable templates for briefs, feature plans, reviews, and docs. Works across all terminal tools.

  • Files in AI-toolkit/prompts/
  • Reference as @plan_feature.md
  • Keep originals versioned

Session Commands

Slash commands for session logging, status, and summaries. Required for repeatable collaboration.

  • Files in AI-toolkit/session-commands/
  • Create .claude/sessions/.current-session

Docs and Templates

Baseline briefs, workflow reminders, and ready-to-edit markdown references bundled inside the toolkit.

  • Claude and Gemini briefing templates for rapid /init
  • Session summary scaffold for consistent logging
  • Workflow and troubleshooting notes you can tailor

Search this page

Need a specific capability? Use the quick search below to highlight sections. For full text search use your browser shortcut (Ctrl+F or Cmd+F).

Type keywords to search agents, prompts, session commands, and docs across the toolkit.

Five step flywheel

  1. Plan - run @create_brief.md for project framing and @plan_feature.md for each task.
  2. Staff - use project-analyst.md and tech-lead-orchestrator.md to select agents per tool.
  3. Build - engage framework agents inside Claude, Gemini, Codex, OpenCode, or Cursor while logging updates.
  4. Review - execute @code_review.md, code-reviewer.md, and brutal-critic.md to catch gaps.
  5. Document - finish with @write_docs.md, /project:session-end, and update session-summary.md.
Keep the workflow checklist from the docs bundle pinned in your editor so every team member follows the same loop.