Devran AI Kit¶
Welcome to the Devran AI Kit documentation — a Trust-Grade AI development framework.
What is Devran AI Kit?¶
Devran AI Kit transforms your IDE into a virtual engineering team with:
| Feature | Count | Description |
|---|---|---|
| 🤖 AI Agents | 26 | Specialized roles (Mobile, DevOps, Database, Security, PR Review...) |
| 🛠️ Skills | 44 | Domain knowledge modules (API, Testing, PR Toolkit, Docker, Research Methodology...) |
| ⌨️ Commands | 40 | Slash commands for every development workflow |
| 🔄 Workflows | 25 | Process templates (/create, /debug, /deploy, /implement, /pr, /pr-merge...) |
| ✅ Checklists | 4 | Quality gates (session-start, session-end, pre-commit, task-complete) |
| 🔗 Hooks | 9 | Event-driven automation (runtime + git-hook enforcement) |
| ⚙️ Runtime | 43 | Runtime engine modules (governance, reputation, self-healing, command bridge...) |
| ⚖️ Rules | 15 | Governance constraints (security, performance, accessibility, data-privacy, testing, git, docs, and more) |
Quick Links¶
NPX install in 30 seconds
26 specialized agents for delegation
40 slash commands for quick execution
44 domain expertise modules
25 full development workflows
Trust-Grade operating constraints
43 modules powering the runtime
Control Claude Code from your phone
End-to-end project lifecycle
7 IDEs from one manifest
Core Philosophy¶
Trust > Optimization. Safety > Growth. Explainability > Performance.
This isn't just a collection of prompts. It's an engineered framework that enforces professional standards through immutable operating constraints.
Key Features¶
- 🔒 Trust-Grade Governance — 15 rules (Performance, Accessibility, Data Privacy + 12 more) with Rigor Profiles (strict/standard/minimal), Scope Filters, and Ethics Gates across all 25 workflows
- 🤖 Multi-Agent System — 26 specialized agents with Instinct System (confidence-scored pattern memory, auto-applied at ≥70 confidence)
- ⚙️ Runtime Engine — 43 modules enforcing workflow transitions, task governance, agent reputation, and self-healing
- 📦 Context as Artifact — Persistent markdown files for plans, specs, and decisions
- 🔄 Continuous Learning — PAAL cycle extracts patterns from every session; Instinct System makes them permanent
- 🛡️ Security First — Built-in secret detection, OWASP Top 10 checklist, vulnerability scanning, and compliance checks
- 🖥️ Cross-IDE Support — Governance configs for Claude Code, Antigravity, Cursor, OpenCode, and Codex; slash command bridges for Claude Code, Cursor, OpenCode, VS Code Copilot, and Windsurf
Release History¶
Latest (v5.2.9): New ui-design-taste plugin adds 5 skills for premium UI/animation work (emil-design-eng, impeccable, taste, animate, improve-animations), wired into the frontend domain rule so they load automatically on UI/animation/design keywords. 1037 tests passing.
v5.2.7: kit update now runs the full gitignore pipeline — projects upgraded from older Kit versions get missing .cursor/commands/, .opencode/commands/, and other bridge entries auto-fixed.
v5.2.6: Fixed Claude Code CLI slash command discovery — blanket .claude/ gitignore narrowed to .claude/commands/ for proper directory discovery. New narrowBlanketClaudeIgnore() migration function.
v5.2.3: Automatic Worktree Support — .worktreeinclude generation for Claude Code, post-checkout git hook for manual worktrees.
v5.2.0: Universal Slash Command Bridge Generation — IDE-native / command bridges for Claude Code, Cursor, OpenCode, VS Code Copilot, and Windsurf with auto-detection and provenance-based safe overwrite.
See the full CHANGELOG for detailed release notes.