AI Dev Workflow

Reusable AI-assisted development workflow, prompts, standards, and documentation templates.

Purpose

This repository contains general-purpose prompts, standards, and templates for AI-assisted software development.

It is designed for implementation-focused work with coding agents such as Cline, Codex, Claude, or similar tools.

What Belongs Here

• reusable coding-agent prompts;
• project bootstrap prompts;
• project planning and design prompts (AI-first architecture, module contracts);
• project development profiles (profiles/) scaling process to project size;
• workflow mode definitions (direct, guided, hybrid, architect-gated);
• smoke-test driven development workflow;
• Git and commit standards;
• deployment workflow: local → git → Forgejo Actions → Docker Registry → server;
• documentation standards;
• handoff standards;
• security review prompts;
• operational baseline materials for logging, audit logging, backup, restore, and readiness review;
• project documentation templates with a repo-local Project Context Layer under docs/context/;
• AI context pack standard and template for agent onboarding and task routing (docs/ai_context/);
• AI-first architecture, module boundary, and cost-aware AI development standards and templates.
• AI-use governance routing for development tools, runtime AI, internal cloud/local use, and hybrid sensitive-data workflows, with generated project-local classification and control records.

What Does Not Belong Here

• secrets;
• client-specific information;
• private infrastructure details;
• company-internal strategy;
• project-specific current state;
• personal memory.

Getting started

See QUICK-START.md for a step-by-step guide to setting up a new project, a common-task reference table, and a summary of the deploy flow.

AI agents should read AI_CONTEXT.md first. It is a short routing map that identifies the correct standards, templates, and prompts for any task type without requiring the full repository to be read.

AI tool entry points are in the repo root: CLAUDE.md (Claude Code), AGENTS.md (Codex), .clinerules (Cline). Project-level versions live in templates/project-documentation-template/.

Recommended Usage

At the beginning of a new project, use the project bootstrap prompt and the project documentation template.

Generated project documentation includes docs/context/ for project-specific AI-readable context. Keep ChatGPT memory for long-term preferences, workstyle hints, and short durable pointers such as: "Detailed project context is stored in the repository under docs/context/." Keep project facts in the repository.

During development, use the implementation prompts and standards as needed.

Structure

• AI_CONTEXT.md - agent entry point: context tiers, task routing, escalation rules. Read this first. Do not read the full repository by default.
• profiles/ - Project Development Profiles scaling process to project size (script-tooling → compact-app → standard-app → modular-app → platform-system).
• global-prompts/ - reusable prompts for coding agents, including a project-design/ category for AI-first architecture and module design.
• standards/ - shared rules for implementation, docs, git, deployment, review, handoff, and AI context pack maintenance; includes planning/design standards, project profile selection, and workflow mode definitions.
• templates/ - copyable project documentation and workflow templates, including project-design-template/ and module-contract-template/ for AI-first design.
• templates/ai-governance-template/ - classification, system/data/control, model, operational, high-risk, transparency, and hybrid reporting records.
• examples/ - small examples showing expected usage.

Operational baseline materials are included for logging, audit logging, backup, restore, and operational readiness.

Deployment materials cover the full flow from local development to Forgejo Actions CI/CD, Docker Registry, SSH-based server deploy, smoke test, and rollback. Use templates/forgejo-actions/ and templates/project-ops/ as starting points for any new deployable project.