Self-Development Organism

Multiple cloud AI agents can collaborate on one codebase through a single shared git repository. Selfdev generates increment-driven prompts and stores progress in git state, so each new agent resumes exactly where the previous one stopped.

What's New: VS Code Agents

Selfdev now ships with four specialized VS Code agents in .github/agents/. Each agent has a dedicated role in the development lifecycle — pick the right one for the task at hand.

VS Code Agents panel showing selfdev agents: do-agent, know-agent, plan-agent, selfdev-agent

know-agent

Knowledge analyst. Scans principles/, cross-references with requirements/ and source code, validates consistency, and produces a structured quality report — preparing the ground for selfdev-agent.

selfdev-agent

End-to-end development agent. Combines plan-agent analysis with do-agent execution in a continuous loop: creates the requirement, implements each increment, runs tests, commits, pushes, and repeats until all increments ship.

plan-agent sub-agent

Planning and analysis sub-agent of selfdev-agent. Clarifies the task, reviews git history for patterns, analyses the codebase, and creates a detailed execution plan as a new requirements/ increment with acceptance criteria.

do-agent sub-agent

Strict execution sub-agent of selfdev-agent. Picks up the current _todo_ increment and implements it exactly as specified — no deviations, no scope creep. Runs tests, commits with traceability, and halts on any ambiguity for approval.

Use Cases

AWS SA Pro Quiz Trainer - Dashboard AWS SA Pro Quiz Trainer - Progress Tracking AWS SA Pro Quiz Trainer - Interactive Mind Map AWS SA Pro Quiz Trainer - Flashcards and Quiz

AWS SA Pro Quiz Trainer aws-cert.trushin.vip

100% Free & No Registration. A comprehensive study application featuring Smart Quizzes with Bayesian Adaptive Engine, Interactive Mind Map with unseen/weak-area filters, Spaced-Repetition Flashcards (SM-2), Service Mastery Heatmaps with completion projection, near-miss scoring, detailed per-option feedback, and built-in progression limits for optimal retention.

System Design Quiz Trainer - Dashboard System Design Quiz Trainer - Interactive Mind Map System Design Quiz Trainer - Quiz Explanations

System Design Quiz Trainer system-design.trushin.vip

100% Free & No Registration. A comprehensive study application focused on Privacy-First learning (Data Stays Local). Highlights include Adaptive Smart Quizzes, Spaced Repetition Flashcards, an Interactive Mind Map with topic filtering, and detailed Explanations with Architectural Diagrams for deeper comprehension.

Wash My Car — Best day spotlight with score, weather breakdown, and recommended wash window Wash My Car — 14-day projection stream with color-coded gauge cards

Wash My Car wash-my-car.trushin.vip

100% Free & No Registration. 14-day car wash calendar powered by weather analysis. Scores each day 0–100 based on rain, wind, UV, humidity & temperature. Best-time window recommendation, post-wash rain outlook, and auto-detect location — all privacy-first with data staying local.

Screenshot of trushin.vip — a portfolio website built with selfdev

trushin.vip

Portfolio website, 100% created by selfdev logic.

Screenshot of cloud-agents.trushin.vip — a how-to guide on Cloud Agents built with selfdev

cloud-agents.trushin.vip

The how-to guide on Cloud Agents.

/* this website */

selfdev.trushin.vip

This website, developed by selfdev logic.

Why Selfdev

Zero Config

Drop the selfdev/ folder into any project and run. No dependencies beyond Python 3.8+.

Increment-Driven

Requirements split into sequential increments with acceptance criteria. Always know exactly what to do next.

Multi-Perspective

Five built-in perspectives (User, Test, System, Analytics, Debug) score your codebase fitness from 0.0 to 1.0.

Any Agent, Any Cloud

GitHub Copilot, Claude, ChatGPT, Gemini, and Cursor can all evolve the same project via one repo. The project state lives in git, not in any single agent session.

Evolutionary Tracking

Every completed increment advances a generation. Fitness history is persisted so you can see progress over time.

Extensible

Add custom perspectives by subclassing PerspectiveAnalyzer. Adapt the system for any domain.

How to Use

One-liner for AI Agents (copy-paste)
Clone https://github.com/evgeny-trushin/selfdev to a temp dir, copy its selfdev/, requirements/, and principles/ folders into my project root, then run ./selfdev/develop.sh. Read the generated prompt, implement it, commit, and run ./selfdev/develop.sh again. Repeat until all increments are completed.

Requirements: Python 3.8+ (no pip packages needed).

Manual Workflow

./selfdev/develop.sh ↓ Read the prompt (increment + principles) ↓ Implement the requirement ↓ git add -A && git commit && git push ↓ ./selfdev/develop.sh ← loop back (verifies done → shows next increment)

CLI Reference

CommandDescription
./selfdev/develop.shShow current increment (default)
./selfdev/develop.sh --advanceMark current done, show next
./selfdev/develop.sh --statePrint organism state & fitness
./selfdev/develop.sh --userDocumentation, UX, features
./selfdev/develop.sh --testCoverage, missing tests
./selfdev/develop.sh --systemComplexity, coupling, architecture
./selfdev/develop.sh --analyticsTrends, patterns over time
./selfdev/develop.sh --debugTODOs, FIXMEs, bugs
./selfdev/develop.sh --allRun all five perspectives
./selfdev/develop.sh --selfdevAnalyze selfdev itself
./selfdev/develop.sh --root DIRAnalyze a different directory

How It Works

Increment-Driven Loop

Requirements are pre-split into numbered increments stored in requirements/. Each increment file contains a description, acceptance criteria, and links to applicable principles.

When you run ./selfdev/develop.sh:

  1. Find the lowest-numbered _todo_ increment file.
  2. Resolve all referenced principles and inject their full text into the prompt.
  3. Output a complete, self-contained prompt: requirement + acceptance criteria + principles + commit message.
  4. You (or your AI agent) implement it and commit.
  5. On the next run, the system detects the commit, renames _todo__done_, advances the generation, and shows the next increment.

Cross-Cloud Agent Handoff

Selfdev enables stateless collaboration across cloud environments: one agent commits and pushes, the next agent clones and continues immediately without chat history transfer.

Agent A (Copilot in VS Code) -> commit increment 0015 -> push Agent B (Claude or ChatGPT in cloud) -> clone same repo -> ./selfdev/develop.sh -> receives increment 0016 prompt Git commit history + organism_state.json = shared project memory

Multi-Perspective Analysis

PerspectiveAnalyzes
UserREADME quality, docs, feature completeness
TestTest coverage ratio, missing test files
SystemMcCabe complexity, file size, coupling
AnalyticsCommit trends, fitness history, patterns
DebugTODOs, FIXMEs, syntax errors, uncommitted changes

Project Structure

your-project/ ├── organism_state.json # Evolutionary state (auto-managed) ├── principles/ # Development principles ├── requirements/ # Increment files (todo ↔ done) └── selfdev/ ├── develop.sh # Entry point (bash) ├── organism.py # Orchestrator ├── models.py # Data models & constants ├── analyzers.py # Code & git analysis ├── perspectives.py # Test & system perspectives ├── user_perspective.py # User perspective ├── diagnostics.py # Analytics & debug ├── formatters.py # Output formatting └── increment_tracker.py# Increment lifecycle